ES3009577A1 - Method and system for improving safety in railway traffic (Machine-translation by Google Translate, not legally binding) - Google Patents

Abstract

Method and system for improving safety in railway traffic. The present invention proposes a method and system for improving safety in railway systems. To this end, data is acquired from various sources (e.g., from infrastructure command and control systems, train systems, train driving personnel, as well as contextual, operational, and other specific data collected by imaging devices and sensors, etc.). All of this data is synchronized and precisely geolocated, and computational analyses are performed from it to generate results that improve safety (e.g., safe driving patterns, detection of risk events, evaluation of the driving experience, detection of complex driving scenarios, among others). (Machine-translation by Google Translate, not legally binding)

Description

DESCRIPTION

Method and system for improving safety in railway traffic

TECHNICAL FIELD OF THE INVENTION

The present invention relates to the field of railway system safety and, more specifically, to a method and system for improving safety in railway traffic through, among other actions, data acquisition, geolocation and synchronization of records from different sources, automatic risk identification and optimal driving patterns, and calculation of safety indicators. This information will be used to implement appropriate measures to improve safety.

BACKGROUND OF THE INVENTION

Operational safety is one of the fundamental characteristics of railway transport systems and is present in all their components, from the lines to the rolling stock that circulates on them, and includes the personnel, as well as the rules and procedures that govern their operation.

Traditionally, risk analysis and monitoring and analysis systems have focused on the study of technical aspects and standard values of reliability, availability, maintainability and safety (better known by its acronym in English "RAMS"), FMEA analysis (Failure Modes Effects Analysis), fault trees, etc. of the different subsystems that make up the railway system (for example, signaling, communications, electrification, etc.). However, in recent years, the analysis of other aspects of safety (such as human factors, organizational factors, critical points of the railway route...) has become increasingly important, given that a large percentage of railway accidents are a consequence of them and, therefore, new regulatory and legislative frameworks increasingly introduce the obligation to consider and monitor them within the safety management systems of both infrastructure managers and railway companies.

Furthermore, many of the records and logs from rolling stock are not synchronized and/or lack precise georeferencing, which would allow for correct geolocation in areas with low GNSS (Global Navigation Satellite System) coverage or to distinguish the track being traveled on when there are several parallel tracks within a short distance.

There are existing solutions for security analysis, but they do not solve the existing problems or take into account the aforementioned aspects and features. For example, patent application EP3654675 describes a system for analyzing data from Chinese CTCS-3 signaling systems (similar to ERTMS, the European Rail Traffic Management System, Level 2), which collects data from the on-board signaling system and its corresponding legal recorder unit (JRU), and data from the GSM-R communications network, among others. This system is focused on monitoring and improving the maintainability and availability of a specific CTCS-3 signaling system with a GSM-R-based communications network. This invention does not perform safety analysis taking into account the human factor in the operation of railway systems, is not applicable to other types of railway signaling systems and does not use other types of data such as those from other train systems, sensors, imaging devices, voice or weather recordings, among others.

Patent application CN107256219 describes a system for storing, synchronizing, and analyzing data from recorders in automatic train control systems. The analysis system used in this invention is focused on detecting and predicting system failures and not on analyzing the human factor in the operation of railway systems. This system does not use other types of data, such as those from other train systems, sensors, imaging devices, voice recordings, or meteorological data, among others.

Patent application CN1165012441 describes a system for centralizing and storing data from legal recorders (JRUs) to a ground monitoring system, which in turn allows for the decoding and analysis of the records. This system focuses solely on providing a storage solution on a ground server that centralizes and continuously stores logs from the different legal recorders of the rolling stock. The system does not synchronize or georeference records from subsystems of the same train, does not use other data sources, and does not focus on the analysis of safety and the human factor in operation.

Given the challenges with existing solutions, there is a need for a procedure that allows for data acquisition and proper processing to perform an analysis of the operational safety of railway systems, taking into account all the aforementioned aspects and characteristics.

SUMMARY OF THE INVENTION

The present invention proposes an improved method and system for data acquisition and processing to improve safety in railway systems.

To this end, the proposed solution involves acquiring data from various sources (e.g., infrastructure command and control systems, rolling stock systems, as well as contextual, operational, and other specific data collected by imaging devices and sensors). This data (pre-processed or raw) is securely sent to a data center (called a data processing and analysis center), where a high-precision synchronization and geolocation process of the events recorded by the different systems is performed, including those not synchronized by the same clock. Once the data is synchronized and geolocated, a computational analysis method is performed, and the data is stored. The results obtained can be represented through a human-machine interface on different types of physical media.

Specifically, in a first aspect of this invention, a method of data acquisition and processing is presented to improve safety in railway traffic, where the method comprises the following steps.

(a) Receive (for example, in a data processing and analysis centre), via one or more communication networks, records of data relating to railway traffic from different data sources (for example, data containing information relating to the traffic of a particular train or a combination of trains and the infrastructure on which that train or combination of trains runs), some of which data sources are located on a train (and other sources may be located, for example, in a data centre of a railway operator);

(b) Checking whether data records received from different sources are synchronized with each other, and if not (i.e. differences are detected between the time information of records from different sources), synchronizing data records relating to the same train, coming from different data sources, based on the identification of recorded events (affecting that train) common to data records from different data sources; (c) Checking the degree of coincidence between information relating to beacons included in received data records and a stored beacon map;

d) Geolocate received data records (from, for example, trains in circulation) using the following steps:

d1) If said degree of coincidence is greater than a pre-established threshold: geolocate the data records that include reference to at least one beacon, based on the location of said at least one beacon on the beacon map and, geolocate the data records that do not include reference to a beacon, based on their time difference with data records from the same train that do include reference to a beacon (since all the train data records are synchronized with each other, thanks to step b));

d2) If said degree of coincidence is less than the pre-established threshold: geolocate those data records that include location information obtained from a GNSS system, from said location information and, geolocate those data records that do not include location information obtained from a GNSS system, from their time difference with data records from the same train that do include location information obtained from a GNSS system (since all data records are synchronized with each other, thanks to step b)).

In one embodiment, one of the data sources in step a) is a mobile mapping system that, using images captured by one or more cameras and/or point clouds captured by one or more laser scanners, of a section of track on which one or more trains circulate, identifies elements of said section of track (using, for example, a statistical classifier based on neural networks, previously trained with images and point clouds of representations of track elements).

The beacon map can be generated, for each track section, from data collected from the recorders of one or more trains while traveling on that track section and information relating to beacons identified by the mobile mapping system on that track section.

In one embodiment, some of the data records received in step a) contain information from the infrastructure signaling system, which is obtained non-intrusively by following the steps below:

- Obtain, by means of one or more video capture devices, images of screens from an infrastructure control center (for example, images from the operator's station or from the video screen of the traffic control center);

- By analyzing the images obtained, recognize by means of an electronic processor the signaling elements included in said images and obtain the geographical position they occupy in the monitored railway installation (for example, by reading the identification of each element) using, for example, a statistical classifier trained with previous images from the control center;

- By analyzing the images obtained, detect, using an electronic processor, the color changes that occur in the pixels that form each signaling element recognized in the image, to correctly identify the state of the signaling element (using, for example, another statistical classifier trained with previous images from the control center).

Data records, once synchronized and geolocated, can be stored on one or more memory units.

From the data records received relating to the driving of a train or group of trains (step a)), once synchronized and geolocated (steps b) and c)), it is possible, for example, by means of an electronic processor to identify one or more data patterns that result in safe driving and/or to identify those data patterns that result in an undesired driving event that poses a driving risk (accidents, excessive speed, delays, etc., are known as driving risk patterns). The identified patterns and events can be stored in an electronic storage unit, for example, together with a score representing the level of safety risk.

Furthermore, driving on a given route and on a given train can be assessed by comparing received data records relating to driving on said route and train with previously generated safe driving data patterns and/or driving risk patterns (generated following the procedure described in the preceding paragraph or patterns defined by users from electronic user devices); based on this assessment, safety indicators relating to the driving of said train on said route can be generated. For this assessment, a statistical classifier based on neural networks previously trained with the safe driving patterns or driving risk patterns can be used.

From the data records received in step a), once synchronized and geolocated (steps b) and c)), relating to the driving of a group of trains, sets of data records involving a complex driving scenario can also be identified. The input data taken into account for the identification of complex driving scenarios may be one or more of the following: geographical data, type of rolling stock, composition, allocation, experience and training of driving personnel, operational core, condition and type of infrastructure, weather conditions, and dates.

To identify a complex driving scenario, a statistical classifier based on neural networks can be used to analyze the input data received over a period of time (once synchronized and geolocated) and/or the distribution of safe driving patterns or driving risk patterns generated.

In one embodiment, some of the data records received in step a) contain infrastructure information obtained non-intrusively by following the steps below:

- Acquire voice recordings of radio communications between traffic management posts and driving personnel;

- From the analysis of voice recordings, in an electronic voice recognition processor, identify the words contained in said voice recordings (for example, by means of a statistical voice recognition classifier based on neural networks, previously trained in the language in which the recordings are made as well as in the phonemes and specific vocabulary used in the conversations);

- Identify infrastructure information and events relevant to traffic by applying search criteria (e.g., event type, train number, registration number, station names, signal names, driving mode, locations, places, etc.) to the words recognized (identified) in said voice recordings.

Some of the data records received in step a) may come from cameras in the driver's cabin of a train, sensors installed on the train or wearables (sensors worn by the train's driving personnel) and, from these data, behaviors or health status of the driving personnel or events in the cabin or on the track on which the train is running, which may affect the safety of driving, may be identified/detected (for example, by using one or more statistical classifiers based on neural networks, previously trained for their identification/detection).

The data sources from which data records are received in step a) can be two or more of the following:

- Data from recording systems existing on a train or group of trains running on the railway system;

- Data from cameras in the driver's cab, sensors installed on the train, or wearables used by driving personnel;

- Data from the railway operator's data centres;

- Meteorological data;

- Data from social networks, fire detection systems or others;

- Infrastructure data, derived from infrastructure signalling systems, from the voice-radio communication system used between traffic management posts and driving personnel, or provided directly by the infrastructure manager;

- Data from a mobile mapping system to identify infrastructure elements.

In one embodiment (periodically or if it is detected that there has been an update to the beacon map in a new area or that the beacon map has changed a lot), the data that had been accurately geolocated by GNSS in step d2) is reprocessed following step d1) to geolocate it from the stored beacon map.

In one embodiment, the beacon map is also updated from data records that include location information obtained from a GNSS system and whose location does not match the location provided in the stored beacon map.

In one embodiment, in step d2) the GNSS system used to geolocate the data records may be a GNSS system external to the train to which the data refers.

In another aspect of this invention, a system is presented for carrying out the methods described above. Specifically, a data acquisition and processing system is presented for improving railway traffic safety, where the system comprises:

- Electronic devices configured to acquire data records from multiple data sources where at least one of said devices is located on a running train and, where said devices comprise communication equipment configured to send (for example, through one or more communication networks), the acquired data records to a data processing and analysis center;

- A data processing and analysis center, comprising communications equipment configured to receive (e.g., via one or more communications networks) the data records, and one or more electronic processors configured to:

- Check whether data records received from different sources are synchronized with each other, and if not, synchronize data records relating to the same train from different data sources based on the identification of recorded events common to data records from different data sources;

- Check the degree of coincidence between the beacon information included in the received data records and a stored beacon map;

- Geolocate data records received through:

- If said degree of coincidence is greater than a pre-established threshold: geolocate those data records that include reference to at least one beacon, based on the location of said at least one beacon on the beacon map and, geolocate those data records that do not include reference to a beacon, based on their time difference with data records from the same train that do include reference to a beacon;

- If said degree of coincidence is less than the pre-established threshold: geolocate those data records that include location information obtained from a GNSS system, from said location information and, geolocate those data records that do not include location information obtained from a GNSS system, from their time difference with data records from the same train that do include location information obtained from a GNSS system.

The scope of the present invention is defined by the appended independent and dependent claims. For a more complete understanding of the invention, its objects, and advantages, descriptive reference may be made to this text and the accompanying drawings.

DESCRIPTION OF THE FIGURES

To complement the description being made and in order to help better understand the characteristics of the invention, in accordance with some preferred examples of practical embodiments thereof, a set of drawings is attached as an integral part of this description, in which, for illustrative and non-limiting purposes, the following has been represented:

Figure 1 shows schematically the operational safety analysis system according to an embodiment of the invention.

Figure 2 shows schematically, in more detail, a specific embodiment of the operational safety analysis system of Figure 1, according to an embodiment of the invention.

Figure 3 shows schematically the different systems and subsystems that are usually found in rolling stock (train), as well as elements used in an embodiment of the invention.

Figure 4 represents the flow diagram followed for the high-precision synchronization and georeferencing of the records and data collected from the rolling stock, according to an embodiment of the present invention.

DETAILED DESCRIPTION OF THE INVENTION

The present invention proposes a method and system for improving safety in railway systems or, more specifically, a method and system for performing a more efficient, accurate and complete analysis/processing of data relating to railway traffic and obtaining results that allow improving railway safety.

Thus, the proposed method and system perform tasks such as record acquisition, decoding, geolocation, and synchronization, automated identification of complex driving scenarios, identification of safe and efficient driving patterns, driving evaluation, and calculation of traffic safety indicators for driving improvement.

The safety analysis carried out allows us to take into account, among other things, the human factor, that is, it takes into account the human intervention that occurs in railway traffic and that affects the safety of railway systems.

A large number of systems and subsystems are involved in railway operations, so the present invention focuses on those that are relevant from a safety analysis perspective, in accordance with the described methodology. Figure 1 schematically shows the operational safety analysis platform (system). In the embodiment shown in this figure, said system comprises a data processing and analysis center (e.g., cloud-based) that receives input data from different sources. This center can be any data center or, more generally speaking, any electronic processor or set of processors (connected to each other) capable of receiving and processing data. The system can also comprise different applications that allow its use from different types of devices.

Figure 1 shows different data sources (100) that can be used (this is only a non-limiting example and the present invention can be implemented without using all of these sources and/or including other data sources). Listed below are some of the groups of data sources that can be used and are shown in Figure 1:

- Data from existing systems in the different types of rolling stock (101); that is, data from the different trains that circulate in the railway system.

- In one embodiment, data from imaging devices, wearable sensors (341) used by driving personnel (343) and other sensors installed in the train (102) are also taken into account. This data source, like the previous one, is located in the rolling stock.

- In addition to the information collected on the rolling stock itself, railway operators collect data so that various computer applications can manage their operations and monitor their internal procedures (this is known as operator systems data, 103). This data may include, for example, assignment codes, fleet positioning, records of potential delays in operations, fleet maintenance status, identification and training status of driving personnel, and incident logs, among others (103).

- Data from external open sources such as meteorological data (104) or data from other open sources such as data from social networks, fire detection systems or others (105) can also be used.

- On the other hand, infrastructure data can be used as input. This data can come from, for example, signaling systems, the voice-radio communication system used between traffic management stations and driving personnel (106). Operational data types generated by other infrastructure management systems can also be used, such as communication of instructions reporting on the status and work being carried out on the road, speed limits and tables, incident logs, or maintenance status, among others (107).

- In addition, the proposed solution uses a system for locating infrastructure elements (called a mobile mapping system, 108), which allows for the acquisition of both georeferenced images with a viewpoint similar to that of the driving personnel (340), as well as generating a map (3D) of the infrastructure itself.

These input data are sent to the data processing and analysis center (116) (which may be a platform deployed in a data center or cloud environment), where an input connector (109) acquires the data from the different sources (100) and transforms them so that they can be consumed by the ingestion computers (110) and (111) depending on the type of data in question, in batches ("batch") or streaming data. In one of the implementations, the input connector (109) can perform a data transformation process directly and/or follow an incremental process depending on the type of source of the input data (100). In some implementations the input connector (109) may also use metadata as well as combine and transform the different input data before sending it to the ingestion containers or computers (110) and (111) or may save it raw (bronze level) or pre-processed and validated (silver level) in the data storage system (112).

Batch ingestion computers (111) perform preprocessing of data in batches. This preprocessing and processing of data follows a logical grouping of activities that perform a task, pipeline, or multiple pipelines. Data can be cleaned, combined, and transformed for further processing (113). The computers (111) also save the information they process (silver level) in the data storage system (112). Streaming computers (110) ingest telemetry data and, in general, data that comes in streaming, and collect, transform, and store it in the database (112).

Depending on the implementation, the data storage system (112) can combine different types of databases, such as relational, non-relational, key-value, graph, columnar, document, etc. The data in turn can be stored in layers as they are processed, validated and transformed in (109), (110), (111) and 113) before reaching an optimized level (gold level) that facilitates efficient analysis and guarantees correct usability of these by the applications (115).

The processing of the data (113) is carried out on different computers that are indicated in more detail in figure 2, which is where the geolocation and synchronization of data and records, automated identification of complex scenarios for driving, identification of safe and efficient driving patterns, evaluation of driving, calculation of traffic safety indicators or generation of a beacon map, among others, are carried out.

The output connector(s) (114) orchestrates and outputs the data in different types of formats so that they can be consumed by the applications (115) and/or other systems (118). In one embodiment, it is contemplated that different computers will be used to create different applications (115) that can be consumed from different user devices (117) such as a web portal, mobile application, desktop application, or a virtual and augmented reality environment. This is only a non-limiting example, in other implementations, the platform (116) may not have the same elements as those listed above and/or may have other/different elements instead of, or in addition to, those listed above.

Figure 2 shows in more detail the different components that make up the operational safety analysis system, taking into account the human factor, according to an embodiment of the present invention.

For greater clarity in the explanation of the present invention, the different systems and subsystems that are usually found in rolling stock (in the case of Figure 3, a train) are shown schematically in Figure 3, as well as elements used in an embodiment of the invention, also mentioned in Figure 2. At the end of the description, the references included in Figure 3 are listed and what each one corresponds to.

The procedure and structures used in the acquisition of data from the different data sources (100) that can be used in the present invention are described below (this is only an example and other data acquisition procedures or structures can be used):

In the data that is collected in the rolling stock (trains), as indicated above, there are two large groups: one is the data from existing systems of the rolling stock (101) that can be acquired by a local computer (206), which is connected in a secure and non-intrusive way to the data bus (201) of the vehicle and forwards the data from the bus. In one embodiment, this data bus can be the data bus of the train control and monitoring system (better known by its acronym in English TCMS, "Train Control and Monitoring System") that controls and monitors the main parameters of the operation of the train. This connection (205) between the computer (206) and the data bus (201) can be adapted to the technology used in the vehicle to be monitored, for example, MVB (Multifunction Vehicle Bus), Ethernet or others.

Additionally, the local computer (206) has a secure and non-intrusive interface that allows downloads (normally periodically or when the user wants) of the different types of data recorders (202, also known as data recorders or storage devices for example) of subsystems that can be found in the rolling stock such as, for example, ASFA system recorders (Automatic Signal Announcement and Braking), ATP systems (Automatic Train Protection), ATO systems (Automatic Train Operation) or energy consumption records among others. In the same way, the system is capable of extracting (for example, periodically) the data from the legal recorder/s, JRUs (203) that may be installed in the vehicles. The computer (206) can have an Ethernet interface with which to communicate and download files from the recorders (202) and JRUs (203) and can have a USB connection that allows it to emulate the file download process carried out by maintenance personnel by using a USB pendrive. The computer (206), once it has the data, sends it securely through a data transport network between the rolling stock and the data center or platform, where the input connector(s) (233) receives, processes and sends it to the computer (234) that is in charge of the rolling stock data, for decoding and processing. In Figure 2, reference 210 designates the equipment (or equipment) of the rolling stock allowing communication via a data transport network (interface or communication equipment of the rolling stock with the communication network) and reference 232 designates the equipment (or equipment) of the data center allowing communication via the data transport network (interface or communication equipment of the data center with the communication network). This transport (communication) network between the rolling stock and the data center may be of any known type (3G, 4G, 5G, etc.) and may in turn comprise one or more communication networks.

On the other hand, depending on the type of rolling stock to be monitored, the system may use one or multiple electronic imaging devices (207), sensors and wearables (208), for example, to monitor the environmental conditions in the driving cabins, the fatigue and attention status of the driving personnel (340), to detect the use of unauthorized mobile devices or any other data of interest that may be captured by images of the train's driving cabin. The electronic imaging devices (for example, cameras) are located in the driving cabins and can acquire images of both the driving personnel's (340) view of the track, as well as of the interior of the driving cabins.

The wearable sensors (341) used by the driving staff (340) allow data to be collected during driving, such as biochemical and physiological variables such as body temperature, blood pressure, oxygen saturation, heart rate, glucose level and physical variables such as accelerations for the detection of possible falls, etc. With this data, an early warning system can be activated in the event of unwanted events that may affect the health of the driving staff and compromise safety, such as feverish situations, drops in blood pressure, respiratory failure, falls, heart attacks, among others, as well as to be able to detect and evaluate stress levels during driving and analyze the situations in which it occurs as part of the processing carried out on the computers (235, 243) of the data center.

These devices (207, 208) can be connected to a local computer (209) that performs local preprocessing of the sensor data and captured images. Just as the local computer (206) did with the data from recorders and the train data bus, the computer (209) that collects the data from the sensors and cameras (207, 208) is connected to secure communications equipment (210) to send the data, through one or more communication networks) to the platform deployed in a data center (also called data center, 116) through communications equipment (232) where the input connector (233) receives, processes and sends them to the computer (235) for processing (for example, for the recognition of data of interest through the images or sensor data received). Figure 2 shows two different local computers (206, 209) but this is just an example and there may be more than 2 local computers; or even, in another embodiment there may be only one local computer that collects and sends all the data from the rolling stock (from 201, 202, 203, 207, 208) to the data center.

The computer (235) can use a statistical classifier based on neural networks, previously trained with images of representations of situations and events on the road that can be viewed from the driving cabin, as well as with images of human gestures and body expressions that allow determining states of fatigue and degrees of attention during driving, from the data collected by the cameras (207). The computer (235) can additionally also use automatic learning ("machine learning"), for example, a statistical classifier based on neural networks, previously trained with biochemical, physiological and physical data that allow determining unwanted events that may affect the health of the driving personnel, as well as traffic safety, from the data collected by the "wearable" sensors (208). These statistical classifiers can provide a score with the probability of correctly identifying the events and situations used in training. If this score exceeds a pre-established safety threshold, the identification of said event in the cabin is determined to be valid (state of fatigue or lack of attention, drops in tension, respiratory insufficiency, falls, heart attacks...) and the result is sent to the computer that synchronizes and geolocates the rolling stock data (243) and/or to the computer that stores the data (112). In turn, from the events identified in the cabin, an alert system that is part of the applications computer (252) can be activated.

In some embodiments, the wearable sensors (341) are connected locally and wirelessly to the computer (209) and in this way data will be collected only when the driving personnel (340) are in the vehicle's driving cabins. However, depending on the implementation, these wearables (341) may have their own communications system that securely sends the data to the platform deployed in a data center (116).

In some embodiments, it is possible to securely send data collected by the computers (206) and (209) through an existing communications network in the vehicle, with its gateway (328) and antenna (329), which allows communication with the data center. In some implementations, the imaging devices (207) may already be installed in the vehicle (331) itself and the local computer (209) be networked with the imaging devices (207) and with (328) and (329) for sending data.

Railway system operators have different computer systems (103) with which they monitor internal operations and processes (obtaining operational data, 211) as well as the real-time positioning of the fleet (212) for example through GNSS systems. As indicated above, these computers that produce operational data (211) and fleet location data (212) can be another possible data source (103) for the present system. This data source is usually located in the data center of the railway system operator. In order to make use of this data, API applications (Application Programming Interface), file sharing and transfer systems or other means can be used depending on the technology used in the systems from which this data comes (211) and (212). The input connector (233) acquires this data through a secure communications network between the communication equipment of the railway system operator (213) and the communication equipment of the data center (232) of the proposed system, preprocesses and saves it and sends it to the computer (243) for processing. If the data center of the operator system and the data center of the system proposed by the present invention are in the same location (or in other words, are part of the same data center) said communication between them can be through a simple network cable.

The present invention also continuously collects open data on meteorological information and warnings (104) made available to citizens by different meteorological information services and public organizations. Through APIs, the data center, through the input connector (233), continuously obtains the data published by the computers of the meteorological information services (214) that is sent through a communication network (communications equipment 215, 232). The input connector (233) receives, processes and sends them to the meteorological data computer (236) that acquires the meteorological stations and warnings/alerts of the areas through which the vehicles circulate.

Similarly, data can be acquired (via input connector 233) from social networks or other open data sources (216). To do this, as in the previous case, API services are used to obtain data that are sent through a communication network (using communications equipment 217, 232). This data is sent to a computer (237) that processes the data from social networks. The input connector (233) acquires information from social networks and other sources by applying search criteria such as keywords, user names, locations, named places, etc.

It should be noted that, although Figure 2 shows separate computers (234-242) for receiving and processing data from each different type of source, this is only one possible embodiment. In other embodiments, there may be one computer that processes data from several different sources, or even a single computer that processes data from all sources. Although the term "computer" is used here, it generally refers to any type of electronic device capable of processing data.

Another source of data for the present invention may be infrastructure data (106) and specifically, for example, data from signaling systems. To this end, a non-intrusive data extraction system may be used from the signaling systems to obtain information on the status of train presence detectors, turnout positions, and signal status, among others. Since this method is non-intrusive, it does not affect the Safety Integrity Level of the signaling facilities. To this end, a video capture computer (218) is used that collects images from signaling systems (for example, images from an operator's station or from the video wall of the centralized traffic control system, CTC) and transforms them into a compressed video format to send them by means of secure communications equipment (221) that sends the data to the data center that receives them by means of also secure communications equipment (232). These images are collected through the input connector (233) and analyzed by a recognition computer (238) of the images captured by the signaling system (and specifically the operator station or the video wall). For the recognition of these images (and the extraction of data of interest) machine learning techniques can be used, such as a statistical classifier based on neural networks, previously trained with images of representations of signaling command and control systems, identifies all the elements of the transmitted image that make up the signaling system. All the signaling elements contained in the images are identified and associated with the geographical position they occupy in the monitored railway installation. In one embodiment, by using another statistical classifier, the color changes that occur in the pixels that make up each element of the signaling system represented in the image are analyzed, to obtain a safety score of the probability of correctly identifying the change in color and state of the signaling element. With this procedure, signaling information can be obtained from images taken from the operator's station, or more specifically, the status of the various signaling elements on the railway installation and their location. This information can be stored in a database.

In addition to the data from the signaling system, in one embodiment the proposed invention can acquire the records of the voice radio communications (219) by means of an interface of a local computer (220) that acquires said voice records. The computer (220) sends said voice records securely by means of a communications equipment (221) to the data center that receives them by means of a communications equipment (232) through a communication network. The input connector (233) receives them and delivers them to a voice recognition computer (239). Said computer (239) can employ "machine learning", for example, a statistical voice recognition classifier based on neural networks, previously trained in the language in which the records are made as well as, in the telephone calls and specific vocabulary used in the conversations between the traffic factors and the driving personnel (340) as well as terms of the railway sector. This statistical classifier can provide a score with the probability of correctly identifying the type of telephone call and the words that make up the voice record; if this score exceeds a pre-established security threshold, a text transcription of the identified words is provided. Additionally, the computer (239) can apply search criteria to these transcriptions, such as keywords, type of event, train number, registration number, station names, signal names, driving mode, locations, places, etc., to identify the communicated actions, as well as events that are relevant from the point of view of traffic safety. In this way, information of interest is extracted from the radio communication records (used between traffic management stations and driving personnel). The results of this voice recognition process can be stored (112) and used in the process carried out by the computer (245) that synchronizes and geolocates the data and associates it with the data from the monitored vehicle.

Additionally, a local computer (224), which may be located together with the interlocking device of the centralized traffic control system, may securely acquire track data from JRU legal recorders (222) and maintenance assistance systems (SAM systems, 223) of signaling installations using the same methodology as that employed by the computer (206) described above. The computer (224) securely sends the data through the corresponding communications equipment (225, 232) where it is collected by the connector (233) and sent to the computer (240) for appropriate decoding and processing.

Infrastructure managers have different computer systems (107), located in the infrastructure manager's data center, with which they monitor operations, the status of the infrastructure and internal processes, recording said data on the status of the tracks in a computer (226). To make use of this data, API applications, file sharing and transfer systems or other means can be used depending on the technology used in the computers where the infrastructure manager records said data (226). These data are sent by means of secure communications equipment (227) to the data center, which receives them by means of communications equipment (232) in an input connector. This preprocesses and saves them (241) and sends them to the computer (245) that performs the synchronization and geolocation of the track data, for processing.

The proposed solution may also include mobile acquisition (which may be periodic) of images by means of imaging devices (228, for example, cameras) and/or point clouds (better known by their English name "points clouds") by means of laser scanner devices (229), of the track layout on which the monitored vehicles circulate. All the records made by these devices are georeferenced with a high precision GSNN system. The acquisition devices (228) and (229) may be located in a train or in another type of vehicle (for example, a car that acquires images of the tracks) and may be connected to a local computer (230) that preprocesses them and sends them securely through communication equipment (231, 232) to the data center, where they are collected by (233) and sent to the mobile mapping computer (242) for processing. The mobile mapping processing computer (242) may employ a neural network-based statistical classifier, pre-trained with images and point clouds of representations of track elements. This statistical classifier may provide a score with the probability of correctly identifying the elements used in training, and if this score exceeds a pre-established safety threshold, the identification of said track element (e.g., a beacon) is determined to be valid.

In one embodiment, part of the image processing performed on the data center computers (235, 238, 242) may be performed on the respective local computers (206, 220, 230) as edge computing to optimize response times and data transmission bandwidth.

In one embodiment, the proposed solution may also allow users (for example, through their devices, 117) to enter information and upload data through the application computer (252) which are stored in one of the storage computers (112) and used in processing (113), such as, for example, infrastructure data (241).

It should be noted that not all embodiments require the use of all the data sources shown in Figures 1 and 2, but only some of them may be used. Thus, in a preferred embodiment, the main data sources used would be:

-101: Rolling stock system data (especially data from recorders and/or JRUs, 202, 203)

-103: Operating system data (211)

-104: Weather data (214)

-106: Infrastructure signaling system data (especially data from the CTC 218 video capture device)

-107: Infrastructure management systems data (226)

-108: Mobile Mapping System data (especially data from the 228 imaging device).

As indicated, the data coming from the vehicle signalling systems and collected in the recorders (202, 203) can come from subsystems that can be found in the rolling stock such as, for example, the European rail control systems (better known by its acronym in English, ETCS, 311), ASFA or another type of automatic train protection system such as the ATP system (302). These data provide information from the beacons located at the foot of the track, which are read by the signalling systems as the train passes by the position of the beacon.

The information from the beacons is captured by means of antennas (305, 306 or 318), normally installed on the underside of the train. The captured information is processed in beacon transmission modules/equipment (better known by their acronyms BTM 304, 319) and processed by the Euro Vital Computer (EVC) system, responsible for the secure processing of the information and for the supervision and control of the speed of the circulation, 303) or by the corresponding process control equipment, ECP (317). These equipment (303) and (317) respectively have odometry subsystems (309, 320) that provide them with high-precision information on the speed at which the vehicle is traveling, as well as the distance traveled since the last beacon read.

In one embodiment, the solution proposed here uses the records of the signaling systems that may be in the vehicle and stored in the recorders, (201, 202 and/or 203), the information of their telegrams (pieces of information sent by a beacon) and the relative distances between beacons, for the generation of a high precision beacon map (244). This beacon map may include, among others, for each beacon: a beacon identifier, type of signaling system to which it belongs, type of beacon, telegrams and aspects of information it provides, relative distance between all of them, geolocation of the beacon, track, kilometer point, direction, link data and name of the beacon group and associated signal if applicable (this is only a non-limiting example and it is not mandatory for the beacon map to include all this information).

The beacon map may also include a track map with the geolocation and kilometer points between beacons. The generation of the beacon map (244) is carried out on a computer (245), which may also use information collected, for example, by the mobile mapping system (108) that provides high-precision geolocation of the identified beacons, and the georeferenced data with GNSS from the rolling stock that is processed on the computer (243). As indicated in Figure 4, the process of generating the beacon map (244) includes an automatic process of identification and validation of possible changes in areas included in the map, such as the installation of new beacons or the elimination of the same, the relocation of existing beacons, modifications to the track that affect the location or distance between beacons, the elimination of beacons or temporary installations of beacons for speed limits.

For generating the marker map, information provided by the infrastructure manager on the location of markers on the roads can also be helpful (although this information from the infrastructure manager is often unavailable or incomplete). Another option is to gather field information on the position of markers, for example, by traveling along the road and using a high-precision GPS system to locate the different markers (although this can be costly in terms of time and resources).

In the embodiment shown in Figure 2, the synchronization and geolocation (using the generated beacon map) of the data from the data sources (101) (102) (103) (104) and (105) would be done on the computer (243). For the synchronization and geolocation of the data, in one embodiment, the flow diagram indicated in Figure 4 is followed (this figure presents a non-limiting example).

These input data (which include, among others, data from recording systems on the rolling stock, imaging devices and sensors) are received (401), for example, via one or more communication networks. These are normally input data from the same train. The first thing that is done is to check whether the data (or rather, the data records) received from the different sources are synchronized (402). If they are not, for those unsynchronized sources (where differences are detected between the time information of their records), a data synchronization is performed (403) based on common events (or more specifically, on the time information of said common events) recorded by the different systems from which the data originate. That is, common events that have left a "trace" in the data from different systems are searched for, the time difference is seen for those data referring to the same common event in the different systems and, once the time difference is known, the data can be synchronized (this common event could be, for example, a train stopping; the time at which said stop is recorded in the different systems is searched for and from there the time difference between them is known and some data can be synchronized with others with a common time reference). The result of this process provides a timestamp to all data and records using the same clock source (for example, from the GNSS system or any other) or time server (for example, the Network Time Protocol - NTP or any other). Once the records are synchronized, the geolocation process begins.

When we talk about a data record, we are referring to a specific piece of data (e.g., train speed) together with one or more pieces of information associated with that data, such as, for example, the location of the data (the place where the data was collected, e.g., where the train was traveling at that speed), temporal information about the data (the time at which the data was collected, e.g., the time at which the train was traveling at that speed), among others. In some cases, such records include references to beacons on the railway infrastructure (e.g., which beacon the data came from or how far from a beacon the data was collected).

To geolocate the data (i.e., determine the location at which each input data item was collected), the first step is to check the information recorded by the train signaling systems (201, 202, and/or 203). If the records contain references to beacon identifiers, as is the case with ERTMS, a search for them is performed directly on the existing beacon map (244). If the records come from a system that does not provide a beacon identifier (i.e., does not reference the data to a specific beacon), the search can be performed on the map using the pattern formed by the relative distances between beacons generated along the route.

In both cases, the level of coincidence of the beacon information in the records collected by the rolling stock with the beacon map is checked (404). That is, the beacon-related information that exists in some of the received data (for example, the relative distances between beacons, the kilometer point at which information about a particular beacon was detected, etc.) is extracted and compared with the existing beacon map. If said information matches that recorded on the map (or rather, if the degree of coincidence is greater than a pre-established first safety threshold, high coincidence threshold), the concordance of the beacon information existing in the data of the records (201, 202 and/or 203) with the information on the beacon map (244) is determined to be valid, and all the data from the rolling stock (405) is georeferenced (geolocated), both the records that include references to the beacons and those that do not, thanks to the fact that all the data is synchronized. In other words, since some of the data is already geolocated because it includes a reference to a beacon (and said beacons are perfectly geolocated thanks to the beacon map), then the other data from the same train are geolocated from this data (even if they do not have beacon information), since all the data is synchronized with each other. It should be noted that this geolocation is not performed using conventional GNSS systems (which, for example, cannot distinguish between roads if they are close together). Rather, as mentioned, the data is geolocated using information on the exact location of a beacon (from the beacon map), which provides much greater precision. In fact, it could be said that the geolocation is performed with odometry precision (odometric precision).

If the beacon-related information in the received data does not match that in the beacon map (the match level is less than the high match threshold), then a determination is made as to whether the match is sufficient or not (406). In other words, a determination is made as to whether the beacon-related information in some of the received data does not entirely match that recorded in the map, but matches most of it (or rather, if the match level is less than the first pre-established certainty threshold but greater than a second pre-established threshold, the medium match threshold). If so, a determination is made as to whether there have been changes to the beacon map (relative to the existing map stored) but that these changes are minor (for example, one or two beacons have been removed or moved, or in general, a small percentage of beacons out of the total beacons). A note is then made as to whether there appear to have been changes to the beacons, but this is not yet considered certain, or, in another embodiment, the beacon map is updated directly. In any case, all received data continues to be geolocated using the same procedure explained in the previous paragraph (405); therefore, in summary, if the match is greater than the first or second threshold, the geolocation is performed using the stored beacon map. In one embodiment, the beacon map is updated as soon as a change in beacons is detected from the received data (410); however, these changes are normally waited for to be confirmed with data from other trains, the beacon map is updated with these changes (410), and from that moment on, the new beacon map is used for geolocation (405).

If the match between the beacon information in the received data and the beacon map is insufficient (i.e., the match level is lower than the second threshold), the stored beacon map is considered invalid for geolocation (this may be because there have been many changes in the beacon positions or because the area where the data was collected does not appear on the beacon map). In this case, an analysis is made to see if the received data contains any records that include GNSS location data (e.g., records from sources 208 or 212 are usually associated with a GNSS location). If so, all data from the rolling stock is geolocated (411), both those records that include its GNSS location and those that do not, since all the data are synchronized. In other words, since some of the data is already geolocated because it includes a GNSS location, the other data from the same train (even if it doesn't have GNSS location) is then used to geolocate the other data coming from the same train, since all the data are synchronized with each other. The geolocation of this data will be done with GNSS precision, which, as indicated, is lower than the geolocation performed from the beacon map (which is odometer-range precision).

The GNSS location of the records may come from GNSS equipment located in the train itself or if there is none, some of the data are geolocated using an external GNSS system (408) (for example, the GNSS system of the operator's data center) and the rest of the data are geolocated using that GNSS information as explained in the previous paragraph (411).

This GNSS information from some of the records can be used to update the beacon map. Furthermore, new high-precision information (413) is generated periodically for the beacon map (e.g., from newly collected input data) and the beacon map to be used for geolocation is updated (405); this information can be used to update the position of existing beacons or to expand the map (because, for example, data is available for an area or a road that was not previously available).

Periodically, or if it is detected that there has been an update to the beacon map in a new area or that the beacon map has changed a lot (412), the data that had been geolocated with GNSS precision are reprocessed (404, 405) to geolocate them from the beacon map, that is, with greater precision.

The computer (245) performs a synchronization of the data from the data sources (106) and (107), which are not located in the rolling stock, with a process equal to that carried out by the computer 243, explained above (steps 402, 403). Likewise, all track signaling elements and points of infrastructure elements relevant to railway traffic can be georeferenced, for example, from data collected by mobile mapping devices (228, 229) or from possible data loaded from user applications that are part of the infrastructure information collected in a computer (241).

With this information, the computer 245 associates the geolocation with the data from these data sources (specifically from devices 222, 223, and 226). The data from the video capture computer (218, which collects images from signaling systems, for example, images from the operator's computer screen or from the video wall of the centralized traffic control system, CTC) can be georeferenced thanks to the analysis performed by the image capture device (238). This occurs because, thanks to this analysis, the location of the signal from which the data was extracted can be determined (for example, by reading the identifier of said signal in the image captured from the CTC, which uniquely identifies its location). Additionally, the transcriptions of the voice recordings of the radio conversations collected by the computer (239) can be geolocated, using keyword searches that can be used to associate the recording with a vehicle and location.

The rolling stock and track data processing computer (246) would be responsible for sequencing and relating the data (once synchronized and geolocated) from all sources; that is, both the data from the rolling stock (trains) and the infrastructure (tracks), as well as other contextual data, and stores them in (112) with a format prepared so that they can be consumed in the most efficient way. For example, by the different applications (115) managed by the computer (252) and by other systems (118), through the output connector (114).

The data, once synchronized and geolocated, can be used to search for safe and efficient driving patterns. This could be done by a processor or computer (247) that would be responsible for searching for and generating safe and unsafe driving patterns from the point of view of traffic safety, taking into account the human factor, as well as efficient from an energy consumption point of view. These patterns may include, for example, speed profiles, monitoring and compliance with traffic regulations, reaction times of driving personnel, or monitoring of processes and procedures, among others. In addition, these patterns may take into account the characteristics and condition of the vehicle, its composition, assignment, weather conditions, among others. This computer can also search for and generate patterns related to degraded situations or events that may pose a risk to safety. For example, by considering input data over a period of time, we would identify those input data that are repeated when there are undesirable events (e.g., accidents, delays, etc.) and/or those input data that are repeated when there is safe and efficient driving (e.g., when these undesirable events do not occur or when driving parameters are within an optimal range). In other words, we would identify patterns of input data (driving personnel behavior, train speed, procedures followed by personnel, infrastructure data) that give rise to undesirable events (which could be called driving risk patterns) and/or those input data patterns that lead to avoiding such events or maintaining driving parameters within predetermined ranges (which we could call safe driving patterns). Thus, with a sufficient amount of data, we would generate patterns (of personnel behavior, speed, procedures) to be followed to improve the safety of rail traffic. The generated patterns and events can be stored (112) along with a score representing the level of security risk.

In one embodiment, based on the collected data, once synchronized and geolocated, the human factor and the efficient driving of the train from which the data was collected could also be evaluated. This would be done by a processor or computer (248) that continuously evaluates the collected data (preferably once processed, for example, in computer 246). To do this, "machine learning" techniques can be used, for example, a statistical classifier based on neural networks, previously trained with the driving patterns and events previously generated (by computer 247 from the collected data or other patterns and events that can be defined by users from their devices). That is, the driving data for each route and for the different trains are compared with the safe and efficient driving patterns generated in the previous section for said route. This statistical classifier provides a score with the probability of correctly identifying the patterns and events used in training in the input driving data (if this score exceeds a pre-established safety threshold, the identification of said event or pattern is determined to be valid). Based on the number of events identified, as well as the evaluation of the degree of deviation between the analyzed records and the reference patterns, risk assessment indicators are generated, as well as whether the train driver's driving is safe and efficient. The data generated in this process can be stored (112). In this computer (248), historical records can be re-evaluated and updated to take into account new patterns and events generated, thereby ensuring uniformity of the evaluation criteria.

In one embodiment, based on the collected data, once synchronized and geolocated, complex driving scenarios could also be identified. This would be done by a processor or computer (249) that massively analyzes the stored data (112). To do this, "machine learning" techniques can be used, for example, a statistical classifier to analyze the input data, the distribution followed by the driving patterns and events that are generated or others that can be defined by users from their devices (117).... Using different filtering factors such as geographic, type of rolling stock, composition, assignment, experience and training of circulation personnel, operational core, degraded situations, condition and type of infrastructure, weather conditions, dates, etc., scenarios can be identified where driving is especially delicate or dangerous, in order to take appropriate measures (or in other words, sets of input data can be identified that imply that driving is dangerous). For example, based on information about where and how traffic signals are run, emergency braking, sudden speed changes, and abnormal driver behavior, among others, it can be determined that a complex/dangerous driving scenario exists in that area under certain conditions (e.g., fog, rain, etc.); and measures can be taken, such as increasing signage, warning drivers, or renovating infrastructure. The identified complex scenarios are stored (112).

Some of the results obtained (for example, the complex scenarios identified, 249, the data from mobile mapping, or the data from georeferenced information, 250) can be transmitted through output connectors (251) to an applications computer (252). This can generate information (for example, scenarios) for driving simulation tools that can be used by users, for example, from mobile (255) or fixed (256) electronic user devices (human machine interface HMI), driving simulators (258) and other systems (259). Specifically, the GIS computer (250) performs all the processing of the relative data of the georeferenced information system (GIS) that, through the output connectors (251), feed the different applications of the computer (252).

The output connectors (251) send the stored data (112) or pipelines from the different ingest computers (110, 111) and processed and transform it into consumable output formats, for example, those that can be accepted by the application computer (252). In one embodiment, depending on the implementation, the output connectors (250) also provide metadata.

In one embodiment, the application computer (252) includes all the components adapted to operate and manage the platform (116) from the different types of devices that the users (117) may use. Communication with the users is carried out through one or more data transport networks (254) (257), which connect to the application computer (252) of the platform (116) through its communications equipment (253). One or more computers can be used to adapt to the different types of devices used for the Human Machine Interface (117), depending on the application (115), such as web display applications, mobile applications, desktop applications, virtual reality or mixed reality applications, etc. The data of the present system may also be shared with other systems (259) through an M2M machine-machine interface so that they can be reused.

Various types of safety, human factor, and energy efficiency indicators may also be calculated (e.g., at the application computer 252). Depending on the implementation, these indicators are adapted to the safety management system of the respective organization. In one embodiment, a real-time alarm system is provided (e.g., at the application computer 252) to send notifications and warnings to various types of users including driving personnel (340) via wearables (341).

In summary, the present invention allows the synchronization of all input data records from different sources, their precise geolocation (with a beacon map) and, based on these synchronized and geolocated data, obtaining results that contribute to the improvement of railway safety, such as, for example, identifying safe driving patterns and risk events, evaluating the driving performed and detecting complex driving scenarios, among others. Based on these obtained results (safe and risk driving patterns, driving evaluation, detection of complex driving scenarios, etc.), appropriate measures can be taken on trains and/or on the infrastructure to improve safety in railway traffic.

In this text, the term "comprises" and its derivatives (such as "comprising", etc.) should not be understood in an exclusive sense, that is, these terms should not be interpreted as excluding the possibility that what is described and defined may include more elements, stages, etc.

Some preferred embodiments of the invention are described in the dependent claims that follow.

Having sufficiently described the nature of the invention, as well as the manner of its implementation in practice, it must be noted that its different parts may be manufactured in a variety of materials, sizes and shapes, and that variations that practice advises may also be introduced into its constitution or procedure, as long as they do not alter the fundamental principle of the present invention.

The description and drawings merely illustrate the principles of the invention. Therefore, it should be appreciated that those skilled in the art will conceive various arrangements which, although not explicitly described or shown herein, nevertheless embody the principles of the invention and are included within its scope. Furthermore, all examples described herein are provided primarily for pedagogical purposes to assist the reader in understanding the principles of the invention and the concepts contributed by the inventor(s) to improve the art, and should be considered as not limiting with respect to such specifically described examples and conditions. Furthermore, everything stated herein relating to the principles, aspects, and embodiments of the invention, as well as the specific examples thereof, encompasses equivalents thereof.

Although the present invention has been described with reference to specific embodiments, it should be understood by those skilled in the art that the foregoing and various other changes, omissions and additions in form and detail thereof may be made without departing from the scope of the invention as defined by the following claims.

References figure 3:

(302) Represents a block diagram of the main elements that make up an ASFA signalling system or other ATP system. It includes:

(316) Human Machine Interface (HMI)

(317) Process control equipment (PCE) and train interface unit (TIU)

(318) ASFA or ATP antenna

(319) BTM beacon transmission module

(320) Odometry subsystem

(202) Registrar

(311) Represents a block diagram of the main elements that make up an ETCS or similar signaling system

(301) Human Machine Interface (HMI)

(303) Euro Vital Computer EVC

(304) BTM beacon transmission module

(305) Euroantenna

(306) Euroradio

(307) Interface unit with the TIU train

(308) STM Module

(309) Odometry subsystem

(310) Antenna

(203) Legal Registrar JRU

(339) Block diagram of the main elements that make up a TCMS vehicle control and monitoring system:

(201) Data bus

(203) Legal Registrar

(321) Human Machine Interface (HMI)

(322) Braking system

(323) Power electronics system

(324) Air production system

(325) Air conditioning system (HVAC)

(326) Other Systems connected to the BUS

(327) Vehicle Control Unit (VCU)

(328) Communications gateway

(329) Antenna

(330) Passenger Information System (PIS)

(331) CCTV System

(332) Other systems

(315) Represents a block diagram of the main elements that make up the voice radio communications system:

(313) Antenna

(314) Voice radio communications

(210) Data network

(312) Data network antenna

Claims (15)

1. Method of data acquisition and processing to improve safety in railway traffic, where the method comprises the following steps. a) Receive (401), via one or more communication networks, records of data relating to railway traffic from different data sources, some of said data sources being located on a train in circulation; b) Checking whether the data records received from different sources are synchronized with each other (402), and if not (403), synchronizing data records relating to the same train, from different data sources, based on the identification of recorded events common to data records from different data sources; c) Check (404) the degree of coincidence between the information relating to beacons included in received data records and a stored beacon map (244); d) Geolocate data records received by following these steps: d1) If said degree of coincidence is greater than a pre-established threshold (405): geolocate the data records that include reference to at least one beacon, from the location of said at least one beacon on the beacon map and, geolocate the data records that do not include reference to a beacon, from their time difference with data records from the same train that do include reference to a beacon; d2) If said degree of coincidence is less than the pre-established threshold (411): geolocate those data records that include location information obtained from a GNSS system, from said location information and, geolocate those data records that do not include location information obtained from a GNSS system, from their time difference with data records from the same train that do include location information obtained from a GNSS system. 2. Method according to any of the preceding claims, wherein one of the data sources of step a) is a mobile mapping system (108) that, based on images captured by one or more cameras (228) and/or point clouds captured by one or more laser scanners, of a section of track on which one or more trains circulate, identifies elements of said section of track, using a statistical classifier based on neural networks, previously trained with images and point clouds of representations of track elements. 3. Method according to claim 2, wherein the beacon map (244) is generated, for each track section, from data collected in the recorders (201,202,203) of one or more trains while traveling on said track section and information relating to beacons identified by the mobile mapping system (108) on said track section. 4. Method according to any of the preceding claims, wherein some of the data records received in step a) contain information from the infrastructure signaling system, which is obtained in a non-intrusive manner by following the following steps: - Obtain, by means of one or more video capture devices (218), images of screens from an infrastructure control center; - By analyzing the images obtained, recognize by means of an electronic processor (238) the signaling elements included in said images and obtain the geographic position they occupy in the monitored railway installation, using a statistical classifier trained with previous images from the control center; - By analyzing the images obtained, detect, using the electronic processor (238), the color changes that occur in the pixels that form each signaling element recognized in the image, to correctly identify the state of the signaling element, using another statistical classifier trained with previous images from the control center. 5. Method according to any of the preceding claims, further comprising the following step performed by at least one electronic processor (247): Identify, from data records received in step a) once synchronized and geolocated, referring to the driving of a train or group of trains, one or more data patterns that result in safe driving and/or identify those data patterns that result in an unwanted driving event of driving risk. 6. Method according to claim 5, further comprising the following step performed by at least one electronic processor (248): - Evaluate driving on a given route and on a given train, comparing received data records, once synchronized and geolocated, referring to driving on said route and said train with previously generated safe driving data patterns and/or driving risk patterns; - Based on this assessment, generate driving safety indicators related to the driving of said train on said route. 7. Method according to claim 6, wherein a statistical classifier based on neural networks previously trained with safe driving patterns or driving risk patterns is used for the evaluation. 8. Method according to any of the preceding claims, wherein the method further comprises the following step performed by at least one electronic processor (249): From data records received in step a), once synchronized and geolocated, referring to the driving of a group of trains, identify sets of data records that involve a complex driving scenario. 9. Method according to claim 8, wherein the input data taken into account for the identification of complex driving scenarios are one or more of the following: geographical data, type of rolling stock, composition, allocation, experience and training of driving personnel, operational core, condition and type of infrastructure, weather conditions and dates. 10. Method according to any of the preceding claims, wherein some of the data records received in step a) contain infrastructure information obtained in a non-intrusive manner by following the following steps: - Acquire voice recordings of radio communications between traffic management posts and driving personnel; - From the analysis of the voice records, in an electronic voice recognition processor (239), identify the words contained in said voice records; - Identify infrastructure information and relevant traffic events by applying search criteria to words recognized in these voice recordings. 11. Method according to any of the preceding claims, wherein some of the data records received in step a) come from cameras (207) in the driver's cabin of a train, sensors installed in the train or wearables (341) used by the driving personnel (343) of the train and where, from these data, behaviors or health status of the driving personnel or events in the cabin or on the track on which the train travels, which may affect the safety of driving, are identified, using one or more statistical classifiers based on neural networks, previously trained. 12. Method according to any of the preceding claims, wherein the data sources from which data records are received in step a) are two or more of the following: - Data (101) from recording systems existing on a train or group of trains circulating on the railway system; - Data from cameras (207) in the driver's cab, sensors installed on the train or wearables (341) used by driving personnel (343); - Data (103) from the data centres of the railway operator; - Meteorological data (104); - Data from social networks, fire detection systems or others (105); - Infrastructure data (106, 107), from the infrastructure signalling systems, from the voice-radio communication system used between traffic management posts and driving personnel, or provided directly by the infrastructure manager; - Data from a mobile mapping system (108) to identify infrastructure elements. 13. Method according to any of the preceding claims, periodically or if it is detected that there has been an update of the beacon map in a new area or that the beacon map has changed a lot (412), the data that had been geolocated with GNSS precision in step d2) are reprocessed (404, 405) following step d1), to geolocate them from the stored beacon map. 14. Method according to any of the preceding claims wherein the beacon map (244) is also updated from data records that include location information obtained from a GNSS system and, whose location does not match the location provided in the stored beacon map. 15. Data acquisition and processing system for improving safety in railway traffic, where the system includes: - Electronic devices configured to acquire data records from several data sources where at least one of said devices is located on a train in circulation and, where said devices comprise communication equipment configured to send, through one or more communication networks, the acquired data records to a data processing and analysis center; - A data processing and analysis center, comprising communications equipment configured to receive, via one or more communications networks, data records and one or more electronic processors configured to: - Checking whether the data records received from different sources are synchronized with each other (402), and if not (403), synchronizing the data records relating to the same train from different data sources based on the identification of recorded events common to data records from different data sources; - Checking the degree of coincidence between the information relating to beacons included in the received data records and a stored beacon map (404); - Geolocate data records received by following these steps: - If said degree of coincidence is greater than a pre-established threshold (405): geolocate those data records that include reference to at least one beacon, from the location of said at least one beacon on the beacon map and, geolocate those data records that do not include reference to a beacon, from their time difference with data records from the same train that do include reference to a beacon; - If said degree of coincidence is less than the pre-established threshold (411): geolocate those data records that include location information obtained from a GNSS system, from said location information and, geolocate those data records that do not include location information obtained from a GNSS system, from their time difference with data records from the same train that do include location information obtained from a GNSS system.