LU102919B1 - Methods and systems for the early detection and evaluation of structural danger spots in road traffic - Google Patents

Abstract

The invention relates to a system or method for the early detection of structural danger spots in road traffic using a digital traffic route network image or map. For this purpose, accident data (1), user input data (3) and sensor data (2), according to their georeferencing, are assigned to segments of the map (500) using a computer system (5). According to one aspect, the computer system (5) carries out an evaluation, comprising at least one frequency determination and/or a data comparison. For the purpose of early detection, a georeferenced segment to which no accident data (1) is assigned is identified as a potential danger point if sensor data (2) or user input data (3) have been assigned with a predetermined frequency and/or if a data comparison, in particular a comparison, has been carried out with an AI-supported model, user input data (3) and/or sensor data (2) shows that there is a predefined degree of feature agreement or feature correlation with critical patterns. This enables early detection of danger spots. According to a further aspect, it is proposed to determine a danger score for identified danger spots.

Description

K720216LU

Pe-Rh March 14, 2022

Applicant:

Initiative for Safe Streets GmbH 53175 Bonn

Methods and systems for the early detection and evaluation of structural danger spots in road traffic

TECHNICAL FIELD

The invention generally relates to methods and systems for

Early detection and assessment of structural danger areas

Road traffic. The invention relates in particular to a computer-aided method for the early detection of structural danger spots in road traffic using a digital traffic route network map using a computer system.

The invention further relates to a computer-aided method for

Determination of a danger score of a georeferenced, structural danger point in a digital traffic route network

Illustration, as well as a computer system for implementing the aforementioned

Process and/or for the use or utilization of the card-like digital data resulting from one of these processes

Traffic route network illustration, which also represents a hazard map.

Furthermore, new uses of the computer system or the hazard map data are suggested. The invention also relates, among other things, to a method or a system for communication with autonomous or semi-autonomous driving vehicles in order to influence the autonomous or semi-autonomous driving behavior depending on danger spots.

A further distinction is made between permanent or structural hazards and only situational, temporary hazards. Permanent

Dangerous areas are always present. Temporary danger spots, on the other hand, are limited in time (e.g. at construction sites) or they only occur in connection with certain general conditions such as adverse weather conditions (be it in the summer due to the glare of the sun, in the rain due to aquaplaning or in the winter due to snow and ice) . There are established solutions for reporting temporary danger spots, such as RDS/TMC according to ISO 14819.

Naturally, early detection is hardly possible here. The present invention relates primarily or exclusively to permanent or structural danger points in the traffic network.

BACKGROUND OF THE INVENTION

Since the number of fatalities in traffic accidents has been stagnating for a long time and no further decline in fatal accidents is expected, intensive research is being carried out into how traffic route networks can be made safer. For some time now, computer-aided methods for identifying people have been increasingly used

Dangerous areas of these traffic route networks are tested and used to tentatively predict accident hazard areas for motor vehicles, typically based on statistical methods or

Probability calculations, and thus to defuse the relevant part of the traffic network before an accident occurs.

STATE OF THE ART

Known methods are based, for example, on using known accident data

To collect and map accident databases from e.g. police authorities and based on the frequency of

Accidents to determine where a danger point exists. With this

approach, danger spots cannot be identified before accidents, some of which are serious, have already occurred. Minor or minor cases are also included in the officially recorded accident data

Accidents without a police report typically do not occur.

There is also the approach of reading sensor data or so-called pulse data through the vehicle electronics and/or additional instruments and using them to detect critical driving maneuvers.

Such pulse data is now used by some insurance companies and/or vehicle fleet operators to monitor the quality of the

to check the driving of the driver and thus to avoid any possible

Be able to adjust premiums. Impulse data can offer an opportunity to identify danger spots through critical driving maneuvers before accidents even occur. Hazardous areas identified in this way can then be, for example,

be reported to road traffic authorities so that they can initiate the defusing of dangerous areas.

However, in practical application it has been shown that

Impulse data as the sole criterion is often not sufficient

To accurately identify or predict danger spots. This is proven, among other things, by the results of a study from 2019, see B. Ryder, A.

Dahlinger et al.; Spatial prediction of traffic accidents with critical driving events - Insights from a nationwide field study;

Transportation Research Part A: Policy and Practice; Vol. 124, 2019 (pp. 611-626), ISSN 0965-8564. This publication shows that although there is a causal relationship between pulse data and

However, there are danger spots, but it also makes it clear that

Pulse data alone is not sufficient to make reliable or accurate statements about danger spots.

Not all pulse data can be traced back to danger spots.

Critical driving maneuvers that are included in the impulse data can be caused by various other causes, e.g. by inattentive drivers, so that there is no connection to the design of a traffic route. Furthermore, in practice it is difficult to detect accumulations of pulse data at certain points on the

traffic network, especially if the accumulations are due to infrastructure, but without any significant danger, e.g. braking in front of traffic lights. Road users with risky and/or inexperienced driving behavior can also be affected

Approaches that rely on sensor data or pulse data trigger false information about danger spots because there is no direct one

Connection between driving behavior and an actual

There must be a danger zone.

A procedure for identifying potential danger spots in the

Road traffic using a central computer unit

Use of a fleet of connected motor vehicles will be in the

DE 10 2020 108 531 Al (Daimler AG). An event that indicates a potential danger point is reported

Pulse data is recorded and sent to the central office with its geoposition

Computing unit transmitted. However, the aforementioned difficulties can only be partially overcome or require increased effort

Computing effort in the vehicles or in the central office

Computer unit. In addition, this solution only applies to motor vehicles.

Furthermore, a procedure for warning of dangers in the

Road traffic and a driving assistance system for warning in the

DE 10 2019 203 405 Al (Audi AG) described. It is also mentioned here that data from various data sources relating to potential danger points should be used for the complex assessment of a potential danger point. Statistical evaluation is mentioned, but how the complex evaluation should be carried out is not described in more detail. In this respect, there is still a need for a solution for detection, in particular

Early detection of danger spots.

A proposal for the complex assessment of danger zones is described on the project homepage of the International Road Assessment Program (iRAP), in particular the publication ~1RAP Methodology Fact Sheet #6 available there. In the iRAP project, the assessment of a danger point is carried out for each 100 m long

Road segment and calculated for one type of road user each.

The factors that are included in the calculation according to iRAP

Risk factors of the road, which indicate how likely it is

Accident is risk factors of the road, which indicate how serious an accident would be,

Speed factors that indicate how the

Probability of accidents depending on the people driving

speed, external factors that indicate how likely it is to be involved in an accident by third parties and a factor that takes into account the probability that a vehicle, for example from oncoming traffic, will be steered onto your own lane. These factors are then used for each

Accident type is determined and the result of the individual accident types are summed up per road segment and road user type, so that a rating is calculated for all accident types for one segment and one road user type. This one on

Probability based approach is used

Early detection of danger spots can only be used to a very limited extent at best. This approach also only targets motor vehicle

perspective.

WO 2017 146 790 Al (Allstate Insurance Co.) in turn describes a system for generating a risk road map, which makes it possible to determine a risk for planned routes.

Previous approaches do not allow for a sufficiently accurate one

Early detection of danger spots, so that, among other things, an unacceptable

A number of false reports of danger zones arise, which is prohibitive for practical usability.

Furthermore, in most cases only the dangers for motorized road users and/or danger spots are taken into account

The view of these road users is recorded, so that there are danger spots for other road users or roads that are primarily for

Motor vehicles are designed for different traffic routes, such as cycle paths or pedestrian paths, not being recorded. Typically, the view of the weaker, non-motorized people is also missing

Road users for potential conflicts.

Another problem is that the traffic route network is only processed incompletely and usually only roughly, broken down into route sections, so that many actual danger spots remain undetected.

Previous approaches also often have the problem that little or no information is available about the type of danger point and only shows where a danger point is located.

TECHNICAL TASK

An object of the present invention is therefore to provide known ones

To at least partially overcome disadvantages or shortcomings of the prior art, and in particular to improve computer-aided methods and systems on the one hand

Early detection and, on the other hand, for the appropriate assessment of

Suggest danger spots. This object is achieved on the one hand by a method according to claim 1 or, independently thereof, on the other hand by a method according to claim 2.

Furthermore, a computer system or a digital hazard map for a computer system according to claim 15 is proposed. Furthermore, for example, according to claim 16, a method for generating hazard-related warning messages to road users is provided

Use of digital hazard maps suggested. More

Uses of the digital hazard maps are proposed in claims 17-19, in particular in connection with autonomous or semi-autonomous driving vehicles, for influencing autonomous or semi-autonomous driving behavior using the digital ones

Hazard cards.

In particular, the procedures and systems should be able to identify structural hazards at an early stage, even before they become apparent through official accident statistics. Of

Furthermore, an object of the present invention can be seen in proposing solutions that make it possible

Transport network as broadly or comprehensively as possible, ideally covering the entire transport network.

GENERAL DESCRIPTION OF THE INVENTION

According to a first independent aspect, a computer-aided method for the early detection of structural danger spots in road traffic or a system set up for this purpose is proposed.

Here, a digital traffic route network map is created using a

Computer system processed, for example map data from the OpenStreetMap (OSM) project or other comparable ones

Map data can be used.

The system or method initially includes providing such a digital map-like image of at least one

Transport network, in particular with infrastructure data and, if necessary,

Traffic route metadata, with the map-like illustration in one

Computer memory of the computer system can be loaded or is being loaded.

The map-like image is first divided into a large number of georeferenced segments of traffic routes by the computer system, the segments preferably being used for the purpose

Data economy and/or more precise localization of

Hazardous areas can have a variable length.

According to a core idea, several different

Data sources, in particular three different data sources, are used, the data of which are appropriately blended by the method or system. Particular preference is given to using digitized georeferenced accident data, georeferenced user input data and georeferenced sensor data that relate to

Movement characteristics of road users relate, in particular to acceleration values.

For this purpose, the invention provides in particular the following steps: a) at least initially providing digitized georeferenced accident data, comprising statistical accident data, in particular in the form of at least one database, which is provided by the

computer system can be accessed or processed; b) Data collection of geo-referenced user input data related to potential danger spots, whereby the data collection is carried out directly or through the computer system, or indirectly through others

Systems or servers can be done that provide the data

provide computer system; and c) data acquisition, in particular by the computer system, of georeferenced sensor data related to movement characteristics of

road users, especially based on

Acceleration measurements that can be generated or are generated in relation to movement by sensors

Road users or vehicles are carried, in particular by vehicle-internal or device-internal sensors, e.g. in known devices such as smartphones,

navigation systems, telematics equipment, IoT devices, etc.;

The temporal frequency or frequency of data collection from the

Steps a) to c) are typically different. The data from the different data sources a), b) and c) can

Each in the form of separate data streams, to different ones

points in time, and if necessary at different time intervals. In particular, no real-time data collection is required, although possible, especially for sensor data c).

If necessary, all data can be standardized by pre-processing the raw data. All or some data can be stored in addition to one

Georeferencing, e.g. GPS coordinates, additional information, such as a time binding or a timestamp.

Statistical accident data for a) can come from different sources

Sources, e.g. from police records and from

Recording by insurance companies.

User input data for b) can initially be collected by a server, e.g. a web server, and updated to the

Computer system are supplied. An ongoing survey is not necessary; short update intervals are desirable. The

Data collection can be initiated by a user and/or by the computer system, for example by querying.

User input data can in particular be collected cumulatively or incrementally. User input data can also be fed directly to the computer system.

The acquisition of sensor data for c) (also called pulse data) takes place in particular, but not necessarily, with short ones

Update intervals, but not in real time, in order to enable the most up-to-date early detection and mapping of danger spots without significant communication effort. The

For example, recording can take place with a time delay or “offline”. The

Recording can take place approximately once a day from smartphones that have an application specific to the use of the invention.

Software, or also from connected vehicles, e.g. via so-called Car2X communication. A collection of sensor data for c) can in particular be at least partially via suitable

Mobile communication takes place, e.g. according to 5G standards, so that data from as many participants as possible.

Vehicles can be detected.

The scope of the invention also includes recording the sensor data via external computers or servers, such as from vehicle manufacturers that collect data from their own fleet via Car2X communication (see, for example, DE 10 2020 108 531 Al or DE 10 2019 203 405 Al).

Neither a direct communication between the computer system and the

Real-time communication is also required between the devices that generate the sensor data. This allows a significant reduction in the

Communication effort and, among other things, anonymization in the sense of

data protection.

Sensor data can be recorded in a filtered manner, for example only in the

Connection with events recognized as critical by the respective device software, or more comprehensively to achieve the most comprehensive possible database, for example for an AI-supported one

Pattern recognition. Sensor data refers in particular to data

Acceleration measurements, which draw conclusions about critical ones

Allow events or maneuvers in road traffic. The

Sensor data for c) can be recorded directly or indirectly by the computer system, e.g. using known IoT techniques or

Interfaces. For acquisition, the sensor data or signals are recorded and digitized using sensors, in particular acceleration sensors, whereby this is done by external devices that are not part of the computer system.

According to a special feature, sensor data for c) is recorded from different types of road users, in particular from motor vehicles (cars, trucks, etc., possibly motorcycles) and at least also from two-wheelers, in particular bicycles and also, for example, e-bikes, e-scooters and the like. , which typically form a more vulnerable type of road user. Therefore are preferred

Sensor data of a first type, namely from sensor values through

Vehicle equipment is generated, and sensor data from a second

Type recorded, which are generated in particular by portable mobile devices, preferably smartphones with appropriate software, so that road users can also use sensors without

Vehicle equipment (e.g. bicycles, e-bikes, e-scooters, etc.) is used. Data collection via different types of

Road users is particularly advantageous, especially for the most comprehensive detection possible, in particular early detection of danger spots, as well as for the validation of data of a different type, e.g. the validation of sensor data of a first type

Type through sensor data of a second type.

Based on the three data categories, the method or system assigns accident data, sensor data and

User input data to segments of the map-like image. This is done in accordance with this data

Georeferencing, such as GPS coordinates or similar

Position information, especially from the computer system.

As a result, affected segments of the map-like image have associated or segment-related recognition data, which, according to the invention, can be used for early detection and/or evaluation

Hazardous areas occur. The segment-related recognition data includes data from at least one or more of the above three

Data categories a) to c), i.e. each include accident data,

Sensor data and/or user input data.

As a result, a large number of the segments have at least assigned sensor data and also user input data or

Accident data. This makes, among other things, cheaper

Amount of training data provided, which enables a more advantageous or more targeted application of AI-supported data processing, in particular machine learning and / or pattern recognition, or a more successful blending of the data using other information technologies. Already in the combination of the three data categories a) to c) for further processing, in particular as training data for machine learning, an independently inventive or inventive feature is seen. Partial registrations, among other things, remain expressly reserved.

Any suitable hardware is available for the computer system used.

Architecture into consideration, a server system, a mainframe computer system set up for AI technologies and, in particular, a cloud

Architecture or e.g. computer clusters.

According to a first aspect, segment-related

Recognition data is then evaluated, at least one frequency determination and / or one

Data comparison, especially e.g. using

Pattern recognition can be carried out.

Segments to which accident data is assigned can be easily done

Recognize danger spots. However, a decisive advantage of the invention lies in the early detection of danger spots for which there is little or no accident data.

Based on the evaluation of segment-related detection data, the computer system can create a georeferenced data for early detection

Segment with detection data that does not contain any accident data, but is nevertheless identified as a potential danger point, is namely determined by the method or system whether i. this segment with a predetermined frequency

User input data and sensor data have been assigned; and/or ii. the data comparison for this segment

User input data and/or sensor data has a predefined one

Degree of feature agreement or feature correlation.

Even a simple frequency determination can

Enable early detection of danger spots, e.g. if a predetermined, e.g. empirically determined first partial frequency of

Sensor data that indicates critical events, with a second one

Partial frequency of user input data, in particular user input data with the same content, is present. Also one

Decision based on a total frequency of

User input data and sensor data is within the scope of the invention.

In a further, advantageous approach, a potential danger point is alternatively/or additionally concluded if a data comparison of user input data and/or sensor data, in particular the sensor data, is sufficient

Shows feature correspondence or feature correlation, especially with segments that have already been recognized as critical and which have been validated, for example, by accident data or user input. Different approaches are used for a suitable data comparison

Consideration, in particular AI-supported data processing, e.g.

Machine learning and/or pattern recognition, e.g. using dynamic

Time warping, support vector machines, random forests and/or artificial neural networks (ANN).

When comparing data, one match criterion can be determined in particular as to whether there is a match, in particular with regard to those recognized by pattern recognition

Categories of information from sensor data from sources of a first type and from sensor data from sources of a second type, e.g. different types of road users, are present.

Sensor data of a first type can in particular include those from

Motor vehicle sensors and sensor data of a second type can in particular be from mobile devices, for example from cyclists.

Based on this, a corresponding one can also be assigned

Relevance information for the danger point under consideration is made. The appropriate use and evaluation of the sensor data from different types of road users is already seen as an independently inventive aspect.

Additionally or alternatively, you can continue to rely on compliance

Information from sensor data and user input data can be checked.

In a further step, the computer system processes all of them as

Segments identified as danger zones, including those for which no accident data is available, are marked in the map-like illustration.

The segments identified as danger spots are preferably assigned a computationally determined danger score, which allows users to display an indication of the level of danger. This can also be done specifically for different types of road users.

According to a second independent aspect, a computer-aided method or system for determining a

Hazard scores of a georeferenced, structural

Danger point in a digital traffic route network mapping, proposed. This can be done in particular on what was described above

System or procedure based.

According to the second aspect, data is collected in particular from geo-referenced user input data related to potential

Dangerous areas, e.g. as explained above.

Furthermore, the aforementioned three data categories a) to c) are used and the georeferenced accident data provided is assigned to the georeferenced recorded data

Sensor data and collected geo-referenced user input data

Segments, each corresponding to this data

Georeferencing. This can be realized in particular by the computer system. As a result, segments of the card-like

Figure segment-related recognition data assigned, each

Include accident data, sensor data and/or user input data.

According to the second aspect, a danger score is computationally determined for segments of the map-like image with associated recognition data by evaluating the data under consideration

Detection data assigned to segment.

In particular, it is provided that - that the danger score of a georeferenced segment with

Recognition data to which no accident data is assigned

Dependence on assigned user input data is determined, in particular validated or relatively increased, for example in

Comparison to segments without associated user input data; and/or - that the danger score depends on a weighting of the

Recognition data is determined using different methods

Weight for accident data, sensor data and user input data, especially with higher weight of accident data compared

User input data and/or higher weight

User input data versus sensor data.

The risk score determined in this way is then assigned or assigned to the corresponding segment. Through the

Use of the above three data categories a) to c) becomes one

Objectification and greater reliability in the evaluation of the

Danger height is only made possible, with the advantage being taken into account that the method or system receives sensor data from different

Road users are recorded and taken into account, so that different danger scores can be determined and displayed for different participants.

The computational determination of a danger score for an identified danger point can be done by calculating, the calculation being a weighting of georeferenced assigned

Sensor data using predefined weighting factors, and preferably at least some georeferenced

Sensor data of a first type, in particular of a first type

Road user type, are weighted depending on a predefined degree of agreement with features of georeferenced assigned user input data and / or further georeferenced sensor data of a second type, in particular coming from a second road user type.

Alternatively or additionally, the danger score for an identified danger point can be calculated by

Processing by means of an AI, in particular using a neural network and/or pattern recognition or the like by the at least one computer system. This can happen in particular under

Event-related characteristics of the infrastructure data from the map image are taken into account.

Beneficial further training, which is the subject of the

Subclaims, but each of which is inventive on its own or is to be regarded as independently relevant to the invention, are discussed below. These are independent for the two above mentioned

Aspects generally applicable, and partial registration remains reserved.

In an advantageous embodiment, data becomes different

Types of road users are obtained, in particular sensor data from different types of road users are recorded.

For this purpose, it is advantageous if the movement-related sensor data includes a type feature that indicates which

Road user type the sensor data comes from and a

Differentiation of at least two different ones

Types of road users, especially motor vehicles and

two-wheelers.

Particularly but not exclusively in combination with movement-related sensor data that have a type feature, an advantageous development provides that at least for

Segments with recognition data for these recognition data, in particular for sensor data with type characteristics, each with a computer pattern recognition, in particular AI-supported

Pattern recognition is carried out and a classification into critical events takes place, in particular through the

Computer system, and the data comparison of user input data and / or sensor data with corresponding data to classify critical events takes place.

To collect safety-relevant events, acceleration data in particular is recorded as sensor data. This can be used to determine the respective trajectory of road users during associated critical driving maneuvers. To

Identification of potentially critical maneuvers can e.g.

Acceleration limits are set. The sensor data can therefore be accessed using a smartphone with a suitable

Smartphone app, a telematics tag, or through IoT devices.

On the one hand, this data can be used for various purposes

Road user types are recorded and, on the other hand, allow

Classify, locate and identify critical situations

to investigate causes. This is based on data from the actual perspective of a road user, e.g

vehicle or two-wheeler.

To classify critical maneuvers of the various

Road users are informed of the calculation methods

Pattern recognition and machine learning are used. Various methods can be used for this, e.g. Dynamic Time Warping, Support Vector Machines, Random

Forests and Neural Networks. Such methods initially make data from different types of road users comparable, which enables analyzes with regard to the

Safety assessment noticeably simplified. Deep learning approaches can also be considered, but are not mandatory, since the data from sources a) to c) are typically already obtained in a structured form, or can be structured through (partially) automatic preprocessing.

To derive conflicts between road users

If danger points are indicated, critical events of all road users at a point under consideration are identified, e.g

Segments superimposed. Based on these findings, a map-based forecast model can be developed, particularly by taking into account infrastructure data and other contextual data

Assessment of critical conflicts can be created. The degree of

Criticality can then be determined through a comparison with officially collected traffic accident data (see data category a) above)) as well as with dangerous situations reported by citizens (see.

Data category b) above) can be determined.

A preferred embodiment therefore provides that, at least for segments with recognition data, this recognition data, in particular sensor data with a type feature, is accompanied by a computer pattern recognition, in particular AI-supported

Pattern recognition is carried out and a classification into critical events takes place, in particular through the

Computer system. Based on this, the data comparison can be carried out

User input data and/or sensor data are used with corresponding data to classify critical events.

According to one aspect, the computer system can be a machine obtained by AI from the three main data sources a) to c).

Learning model, use to improve the map-like representation, in particular to the transport network-related coverage of the

expand early detection. The hazard map can therefore

Provide training data for a model that is used recursively

Self-optimization of the hazard map is used. The expansion can relate to those positions or segments for which there is insufficient data from sources a) to c) or data with insufficient significance. This means that potential ones can be identified based on infrastructure information alone

Hazardous areas can be identified by comparing patterns. In particular, if only sensor data is available, c) can be used in conjunction with

Infrastructure data through pattern comparison with validated ones

Dangerous areas can be identified early. In the

When the machine learning model is created, the accident data a) or user input data b) can be used in particular to validate patterns classified as critical

Optimization of the model can be used.

According to one aspect, for the purpose of early detection, the computer system can provide a georeferenced segment with detection data to which no

Accident data assigned to it can be identified as a potential danger point if a data comparison is carried out for this segment

Sensor data sufficient feature match or

Feature correlation with sensor data of a critical event is used to determine the feature match or

Feature correlation the upstream pattern recognition is used.

Regardless of the type of data comparison used, it can be provided in a data-technically favorable implementation that:

Danger point identified segments are each assigned a data object, which comprises at least one computationally determined danger score, the data object further comprising: - at least one relevance indication for which or which of several different types of road users, in particular pedestrians, cyclists or motor vehicle drivers, one

Danger is imminent, with the relevance information preferably being determined from the associated recognition data; and/or - at least an indication of the type of danger as to which

Traffic behavior, which situation and/or between which

Road users are at risk of danger at the danger point, whereby the

Hazard type information preferably from the assigned ones

recognition data is determined; because preferably they give

In addition to the type of danger, detection data also provides information regarding the specific dangerous situation and, if applicable

Conflicts between road users and/or - at least a contextual indication as to which ones

Environmental influences or traffic situations, at the danger point

Danger threatens, with the context information preferably being determined from the assigned recognition data and/or from additional ones

Traffic route metadata, in particular comprehensive information regarding

Weather dependencies, speed guidelines, time-dependent

Traffic volume or something like that.

It is advantageous if the data objects for danger spots at least indicate which road users are in danger or are at risk (relevance information) and what exactly the danger is (indication of the type of danger). To store the data or as

Any type of data set can be considered as data objects, with the data coming from different sources a) to c). can be brought into a unified format through preprocessing, e.g. in the course of AI-supported pattern recognition or the like.

With regard to data communication, one embodiment provides that data collection from georeferenced

User input data related to potential danger spots continuously and/or sporadically via an Internet interface

Computer system takes place. This Internet interface can be connected to at least one or several servers, in particular web server(s), each of which provides a user interface for user input data. In this way, several possibly different interfaces can be provided, e.g. a classic website and an interface for smartphones.

Apps. Also, for example, an integration of insurance

Accident reports from road users via a corresponding

Interface possible.

According to a further advantageous aspect, the map-like image is divided into a plurality of georeferenced ones

Segments in a variable and possibly iterative manner. Variable subdivision into segments of different geographical lengths can therefore be provided, in particular depending on the transport route infrastructure (e.g. the type of route with longer ones

Segments e.g. on motorways) and/or depending on assigned detection data and/or depending on identified danger spots. This enables, among other things, sufficient local discrimination of danger spots.

Furthermore, the assignment of accident data, sensor data and

User input data for individual segments, determining a geographical distance between the recognition data from them

Include georeferencing so that mapping can be based on relative distances. This can also, among other things, improve local discrimination of danger spots.

Since collecting data from different road users is seen as particularly advantageous, one embodiment for data collection provides that sensor data based on

Acceleration measurements are recorded from different sources, in particular sensor data of a first type, which are from

Sensor values from the vehicle equipment, in particular

Vehicle sensors or telematics equipment are generated by vehicles, in particular motor vehicles, and at least sensor data of a second type, which are generated from sensor values from acceleration sensors in portable mobile devices, in particular smartphones, carried by road users. Other sources are also possible.

The sensor data is preferably provided with time stamps, among other things. The timestamps can be used as a feature

Data comparison with regard to a predefined level

Feature correspondence or feature correlation can be used, in particular as to whether the timestamps are spaced apart in time within a predefined time range or correlate based on the time of day.

According to a further, possibly independently advantageous aspect, the invention provides that georeferenced sensor data is provided by an application-specific smartphone application for detecting potential danger spots, wherein - the smartphone application is preferably set up to detect potentially safety-critical events, in particular on the basis of an exceedance of

Acceleration threshold values, and georeferenced sensor data only for potentially safety-critical events, the sensor data provided preferably

contain information about movement behavior within a time window around the safety-critical event; and/or - the smartphone application provides an interface for georeferenced user input data related to potential

Provides danger spots.

When using a specific application of the invention

Smartphone application for identifying potential

Hazardous areas can further advantageously be provided that - when a potentially safety-critical event is detected, the smartphone application assigns corresponding GPS coordinates for georeferencing, and preferably at least one time stamp, to corresponding sensor data and with the

Sensor data transmitted to the computer system; - the smartphone application, especially when a potentially safety-critical event is detected and/or

Detection of local proximity to a segment identified as a potential danger point, the user to enter

Prompts user input data, including at least one

Confirmation of the existence of a danger zone, preferably also for entering a relevance information regarding affected road user types and/or a

Hazard type information and/or context information and the

User input as user input data with the sensor data

computer system transmitted; and/or - the smartphone application responds to a potentially safety-critical event based on a movement profile determined using an integrated acceleration sensor

Relevance information regarding affected road user types is provided and transmitted to the computer system with the sensor data.

The information, which is preferably assigned to each identified danger point, preferably includes information about the type of hazard

Danger. At least one person can determine this computationally

Hazard type information must be provided, whereby the hazard type information is selected

Information from infrastructure data and/or sensor data and/or accident reports assigned to the segment and/or

User input data, preferably from a combination of

Information is determined from at least two different of these data.

If necessary, danger scores calculated with higher resolution can be divided into a reduced number

Danger levels can be represented more ergonomically. Thus one can

Assigning a danger level, from a number of discrete ones

Danger level for each danger score, whereby

Each danger level has a predefined value range

Danger scores are assigned.

Furthermore, it is possible to determine several computationally

Danger scores for an identified danger point in

Dependence of the respective relevance information is made as to which one or which of several different ones

Types of road users, especially pedestrians, cyclists or

Motor vehicle driver, danger is imminent. This means that participant-specific risk scores can be different

Road user types can be assigned.

In an advantageous implementation, a gradual implementation takes place

Updating the in the digital image of

Traffic network included data, especially the

Danger scores when a predetermined amount of additional sensor data has been collected and/or a predetermined amount of additional

User input data was collected. Thus, it can

Danger zone map can be gradually optimized and continuously updated without requiring significant computing effort, for example for new pattern recognition.

In a variant, it can be provided that the data acquisition of georeferenced sensor data, based on the

Movement behavior of a road user changes in a different way when the local proximity of the road user to a location that has already been identified as a potential danger spot is detected

Segment, the data collection of sensor data can then take place in particular with increased sensitivity or intensity.

This can be done, for example, by suitable programming of a smartphone

Application can be realized and enables an improved

Information on potential danger spots, which in turn enables positive feedback to improve pattern recognition.

According to a further independent aspect, the invention relates to a digital hazard map generated according to one of the embodiments discussed above, i.e. with as

Hazardous areas identified segments, especially by

Early detection identified danger spots where there are none

Accident data is available and/or segments with hazard scores, in particular hazard scores determined on the basis of the three data sources a) to c).

Accordingly, the invention also relates to the use of the resulting digital hazard map (card-like digital

Traffic route network mapping), which according to one of the above

Approaches using georeferenced ones provided by

Accident data, collected geo-referenced user input data and collected geo-referenced sensor data, segments of the traffic routes identified as danger spots.

The invention therefore also relates in particular to a computer system, in particular a cloud computer system, comprising a map-like digital traffic route network image, which was generated according to a method according to one of the embodiments discussed above, and in particular identified as danger spots

Has segments with a danger score.

This system can be used in particular to generate hazard-related warning messages to road users.

For this purpose, terminal devices communicating with the computer system,

For this purpose, terminal devices communicating with the computer system, including terminal devices installed in vehicles (e.g. navigation devices), but in particular smartphones, can issue a warning message, in particular a warning message comprising a, when local proximity to a segment identified as a danger point is detected

Relevance information and/or hazard type information and/or a

Context statement. By sending an appropriate warning message to the

Road users, especially with early detection of

Dangerous areas, overall road safety can be noticeably improved, especially for types of road users indirectly affected by the warning, e.g. pedestrians or

Cyclist.

The resulting hazard map can also be used advantageously in various other applications.

An advantageous application lies, for example, in determining safe route options in one go

Navigation system, in particular in addition to fast and/or ecological route options determined by the navigation system. This can then also take into account previously identified danger spots.

Another advantageous application lies in the optimization of

Solutions for autonomous or semi-autonomous driving. The system can be used for communication, particularly in areas relevant to hazards

Information from the digital hazard map can be used with autonomous or semi-autonomous driving vehicles and, on this basis, allows targeted influence on autonomous or semi-autonomous driving behavior depending on danger spots, in particular taking into account relevance and/or

Hazard type and/or context information and/or to influence route selection. In this way, driving behavior can be adapted automatically, specifically and better to areas with high risk potential.

Another advantageous application is the provision of training data sets for training a machine learning system.

Model for predicting potential danger spots, e.g. for recursive optimization or expansion of coverage in the digital danger map itself.

The resulting machine learning model can also be used in particular in a traffic planning simulation. In this way, improved analysis tools can be implemented for traffic planners and authorities, which already carry out hazard simulation

Enable or significantly improve planning.

BRIEF DESCRIPTION OF THE FIGURES

Further features and advantages of the invention can be found in the following description of a preferred exemplary embodiment based on the accompanying figures, without limiting the scope of protection. Show here:

FIG. 1 a schematic basic concept for data processing, in particular for determining a so-called danger score for a

transport network;

FIG. 2 shows a data flow diagram for data technology

Illustration of an embodiment of the invention

Method for the early detection of danger spots and the computer system according to the invention;

FIG. 3 shows a flowchart or flow chart of a

Embodiment of the method according to the invention for early detection and evaluation of danger spots;

FIG. 4 shows purely exemplary user interface views of an app or website for hazard-related user inputs that are fed to or recorded by the computer system or method; and

FIG. 5 shows an exemplary representation of one

Danger zone map with danger zones indicated based on discrete danger levels.

DETAILED DESCRIPTION BY THE FIGURES

FIG. 1 outlines the underlying concept of a

Embodiment of the invention. This embodiment looks like one

The core feature is the processing and analysis of data from three independent and different data sources for a transport route network-wide identification of danger spots and/or

Hazard score calculation. Become the first data source

Accident data 1, in particular from the police and/or by

Accident data registered by insurance companies is used, which is recorded, for example, by the police authorities via IT systems. This at least covers more serious traffic accidents that have already occurred, which can also affect various road users. From such accident data 1 alone, no statement can be made about dangerous areas where no accident has yet occurred. Furthermore, in the case of minor accidents, the police are often not informed, so that these accident scenes are in the public domain

Statistics do not appear, or the recording of such accidents by authorities is not required.

A resulting unreported number of minor accidents as well

Near misses and other conflict situations should be uncovered via two additional data sources, namely on the one hand

Hazard reports from road users in the form of

User input data 3 as well as sensor data 2 (so-called pulse data), which are supplied from motor vehicles with suitable sensor equipment and/or from smartphones, especially during critical driving maneuvers (e.g. evasion or heavy braking). For the

To generate danger reports from user input data 3, a web platform and/or a smartphone app is provided (see FIG. 4). Here, road users can for the entire

Traffic network danger spots on a digital image of

Report and comment on traffic route networks in the form of a digital, interactive map (see FIG. 4) and those that have already been recorded

Validate, falsify and/or evaluate danger spots. In this way, “crowdsourcing” can be used to identify danger spots throughout the transport network.

FIG. 2 shows a basic data flow diagram, based on which the

Process of an embodiment of a method according to the invention for identifying a danger point in a digital

Mapping of traffic route networks, or the so-called danger map, is explained. The approach is based on that in FIG. 1 shown

Concept. Four data sources are shown schematically here

Form of digitized georeferenced accident data 1, georeferenced sensor data 2, which generate sensors that are carried by road users or vehicles, georeferenced and related to potential danger spots

User input data 3, as well as optional, georeferenced

Contextual data 4. Data from these sources 1, 2, 3, 4 are used by . from at least one computer system, e.g. a cloud computer

System 5 processes, georeferenced in this case means that the data is at least one related to the geographical position

Information is assigned, e.g. GPS coordinates or dal.

The digitized georeferenced accident data 1 is, for example, in

Form of a database 1 provided to the cloud computer system 5,

This georeferenced accident data 1 may have been collected or recorded in advance by authorities, such as road traffic offices and/or police authorities, and/or, for example, by insurance companies and/or vehicle rental companies and made available by them; if necessary, data is processed in a suitable manner for this purpose unified formats.

The georeferenced sensor data 2 are continuously and/or sporadically recorded directly or indirectly by the cloud computer system 5, with this georeferenced sensor data 2 being collected by the

Electronics of vehicles participating in traffic and/or preferably of devices which road users, such as e.g.

Car drivers, cyclists or e-scooter drivers, carry with them, especially smartphones, smartwatches, telematics tags,

Navigation systems, IoT devices, etc. generated. These include

Sensor data 2 movement values, in particular acceleration values, and position information data assigned to them, e.g. GPS coordinates.

The sensor data 2 can be recorded, for example, by one or more servers, not shown, with which the Cioud computer

System 5 communicates. Furthermore, the sensor data 2 can preferably include time information or time stamps, which

Specify the time of data generation. In more advantageous

Embodiment, the sensor data 2 can also include information about the corresponding type of road user, which results inherently from the data source, for example in a vehicle navigation system or a telematics system, or can be supplemented by the same, for example a smartphone APP. Georeferenced sensor data 2 can be continuously, automatically collected by the system.

The georeferenced user input data 3 related to potential danger spots is collected user-initiated or system-initiated, with or without a direct time commitment. The

User input data 3 can also be collected via an intermediate server (not shown), e.g. a web server, with which the cloud computer system 5 communicates. The input can be user input data 3 can be initiated by a user. This user input data 3 includes subjective information about at least one, at least for

Potential danger point perceived by the user as such

Information including geo-referenced position information, e.g. GPS

Coordinates of the potential danger point and/or information regarding the type of danger of the potential danger point and/or

Information regarding the relevance of the potential

Danger point for one or more different

Road user types.

Optional georeferenced context data 4 are automatically recorded continuously and/or sporadically, the context data 4 comprising temporary information, in particular time-dependent information, such as weather data and/or information on the traffic situation and/or information on the season,

The cloud computer system 5 processes the recorded georeferenced data to determine a danger spot

Basis of the georeferences assigned to the data

Position information. If a danger point has been identified, the cloud computer system 5 creates or changes one accordingly

Data record or a corresponding data object in the

Hazard map.

Furthermore, the cloud computer system 5 can trigger geo-referenced warning messages or warnings 7 related to a danger point from terminal devices 6, especially if they are in this area

The danger zone can be geographically approached, received, interpreted and, in particular, displayed visually and/or acoustically for a user, for example. The information from the cloud computer

System 5 can also be used to communicate with autonomous or semi-autonomous driving vehicles in order to influence the autonomous or semi-autonomous driving behavior depending on

Dangerous areas are used. Optionally, warning 7 can be a

Trigger user query, which end devices 6 brings the user after verification, falsification and/or after a

Comment related to the danger zone. The

User response 8 to such a user query can be from

Terminal devices € are transmitted to the cloud computer system 5 for verification, falsification and/or evaluation of the

Dangerous areas. So user input data 3 and/or

User responses 8 are processed by the cloud computer system 5, e.g. for those already identified by sensor data 2

To check danger spots and/or to check those that have already been identified

Modify dangerous areas.

FIG. 3 shows a flowchart of an embodiment of a method according to the invention for identifying and verifying a danger spot using georeferenced information from several different data sources, in particular accident data 1 and sensor data 2 and user input data 3, each if available.

In the process according to FIG. 3, software arranges 301 in a first step

Information from georeferenced detection data, accident data 1 and/or sensor data 2 and/or user input data 3, corresponding to the georeferenced position information, is assigned to a segment of the digital map.

After the assignment of the information follows in a second

Step 302 a comparison of a georeferenced potential

Information assigned to the danger point, in particular

Comparison of information from at least two different ones

Data sources 1, 2, 3. Will be a predefined level

Correspondence of those assigned to a potential danger point

Information, e.g. from at least two of the three different ones

Data sources 1, 2, 3, achieved and/or sufficient

Partial frequencies of critical sensor data 2 and/or

User input data 3 reached and/or when information is out

If accident data 1 is available, you can continue with a third step 303. Since information from accident data 1 can be viewed as officially validated, it generally does not require any verification and can serve as the basis for the

Marking or identification (already known)

Dangerous areas are used.

In the third step 303, a segment of the digital

Hazard map identified as a danger point and provided with corresponding data in the data record or data object, including a relevance statement as to which one

Type of road user is threatened with danger, a danger type information, e.g., which traffic behavior poses a danger, and a context information about which environmental influences or

Traffic situations where danger threatens. To

Obtaining this data can include the recognition data 1, 2, 3 and/or additional traffic route metadata, in particular comprehensive

Information regarding weather dependencies,

Speed guidelines, time-dependent traffic volume, etc. can be used.

In the second step 302 the predefined degree is displayed

Agreement not reached and no relevant ones exist

Accident data 1, a further test step 303° follows. If, for example, the information situation from sensor data 2 or user input data 3 for a segment to which no accident data 1 is assigned is not sufficiently clear, a special data comparison can take place in step 303 '.

In step 303*, in particular, user input data 2 and/or

Sensor data 3 to a predefined level

Feature agreement or feature correlation is examined. The

Computer system 5 can use a georeferenced device for early detection

Identify segment without accident data 1 as a potential danger spot if a data comparison is carried out for this segment

Sensor data 3 sufficient feature match or

Feature correlation with sensor data 3 of a critical event shows what is checked in a further test step 304 '.

As a basis for computational testing

Feature matching or feature correlation can be achieved, for example, by the computer system 5, which is AI-supported in advance

Pattern recognition can be used, due to which a

Classification of all recorded georeferenced detection data 1, 2, 3 into critical events takes place. A technique known per se can be used for pattern recognition, e.g. pattern recognition using dynamic time warping, support vector machines, random

Forests and/or artificial neural networks (ANN).

In step 303 'can be won based on such

Classification can be used to compare sensor data 3 with corresponding data to classify critical events.

If the vote in test step 304' shows that it is sufficient

Feature agreement or feature correlation is present

Segment is again identified as a danger point in the third step 303 and provided with corresponding data. We in

Step 304' no sufficient feature match or

This can be determined by feature correlation, for example

Model optimization can be used in step 305'.

In the third step 303, the referenced potential

Additional information, including e.g. from geo-referenced context data 4, can be assigned to the danger point. Additionally or alternatively, in the third step 303, a user request is generated, which asks at least one user to validate the potential danger point by entering 4.

In a fourth step 304, the data on the segment is saved and an evaluation is carried out by calculating a

Danger Score. For this purpose, the danger score can be calculated by determining it depending on a weighting of the detection data, each with a different weight for

Accident data 1, sensor data 2 and user input data 3, in particular with a higher weight of accident data 1 compared to

User input data 3 and/or higher weight

User input data 3 versus sensor data 1.

The hazard map can then be updated in a fifth step 305.

An important data source for the method according to the invention are

User input data 3. Experiments have shown that the so-called crowdsourcing approach for proactive collection of

Hazardous areas in road traffic can generally be used. On an interactive danger zone map, such as in

FIG. 4 presented in the form of an app and a homepage

Users submit danger reports.

FIG. 4 shows an embodiment of a digital

Danger zone map in an app or a homepage

Generation of user input data 3, which can be used in a method according to the invention (see FIG. 2

Reference number 3). The user has the option of information in

Form of a hazard relevance 401 (relevance statement), one

Danger type 402 (danger type information) and the exact geographical position of the danger point 405 (georeferencing). The danger relevance 401 indicates which ones

Types of road users the danger spot represents a danger,

Furthermore, the danger zone map offers the possibility of storing comment texts 403 and images 404 related to a danger zone, so that 2.B. a danger zone can be described in a way that is understandable for other users, even if they are not in the area, for example, and/or to provide further information for any users

Authorities/offices to be specified. This user input data 3 the

Danger zone maps can then be used, for example, by an embodiment of the method according to the invention according to FIG. 2 from the cloud

Computer system 5 are collected. Furthermore, one offers

Embodiment of a danger zone map offers the possibility of evaluating already identified danger zones for validation and modification and/or requests a user to validate an identified danger zone (see FIG. 2

reference number 8). Marking a danger point

User input on the interactive map, for example, runs in three

Main steps: First, the road user marks the corresponding location on the map and can then choose the type of

Danger determines, for example, whether the place is unclear or has poor road conditions (402). In the second

Step can then be selected for whom the position is

Source of danger: e.g. pedestrians, cyclists, motorcyclists and motor vehicle drivers (car, truck or bus drivers) and thus generates a danger relevance 401 (relevance information). In the next step, the more precise danger trigger can be specified, such as unclear traffic routing or

Wild growth on the side of the road. Furthermore, the app or website (not shown) can provide for the user to have one

Context information as to which environmental influences or

Traffic situations, typically danger threatens at the danger point.

FIG. 5 shows an example of the use of a device according to the invention

Danger level 502 determined by the procedure in one

Danger zone map 500 for an urban area, whereby the determined danger levels 502 are compared to the accident clusters 501 that were determined by the accident commission for the same period. The danger levels are graded in color: the darker, the more dangerous a section or junction is. The accident sites or accident clusters are shown as circles 501. It can be seen that accident sites or accident clusters 501 only cover a part of the road sections and junctions 502 identified as dangerous,

In addition, the danger level 503 varies between 5 levels (from 1 to 5) at the accident frequencies 501. This means that a distinction is not only made between the presence of a dangerous place and the absence of a dangerous place. In this way, the danger score can be used to provide further information about the

Road safety can be determined and presented ergonomically for users. In addition to danger areas that are independent of

Environmental factors exist at all times, there are danger spots that only exist in certain contextual situations, e.g

Heavy rain, at night or when there is high traffic, so that warnings are only issued in the appropriate context situation, the database is filled with additional context data 4 such as weather data and

Traffic data enriched. This means that the risk assessment can be carried out using

Danger score dynamically takes into account the respective context situation and the optimal one for the respective user groups

convey information content. By means of an example and excerpts in FIG. 5, the danger zone map 500 shown can be the results of an inventive

The risk level 502 determined by the procedure can be verified and at the same time show that the accident data 1 in the form of accident clusters (circle) 501 is not sufficient to identify as many dangerous accidents as possible

Identify locations in a traffic route network.

Tests showed that by means of an inventive

The process also determines the danger map, among other things

Early detection of structural dangers in road traffic is suitable.

K720216LU

Pe-Rh March 14, 2022

Applicant:

Initiative for Safe Streets GmbH 53175 Bonn

Procedures and systems for the early detection and evaluation of

Dangerous spots in road traffic

Reference symbol list

FIG.1-3 FIG.4 1 Accident data 401 Hazard relevance 2 Sensor data (relevance information) 3 User input data 402 Type of danger 4 Context data 35 (Type of danger information) 5 Cloud computer system 403 Comment text 6 End devices 404 Image 7 Warning 405 Position of the 8 user response Danger point 40 (georeferencing )

FIG.3 301 first step FIG.5 302 second step 500 danger zone map 303 third step 501 accident cluster 304 fourth step 45 502 road section/ 305 fifth step junction 303' first test step 503 danger level 304' second test step 305' optimization step

Claims (19)

K720216LU PE-RH March 14, 2022 Applicant: Initiative for Safe Streets GmbH 53175 Bonn Methods and systems for early detection and assessment of dangerous spots in road traffic Patent claims 1. Computer-aided method for the early detection of structural danger spots in road traffic using a digital traffic route network image using a computer system, the method comprising the following steps: - Providing a digital map-like image of at least one traffic route network, in particular comprising infrastructure data, the map-like image being converted into a Computer memory of the computer system can be loaded or is being loaded; - dividing the map-like image into a large number of georeferenced segments of traffic routes by the computer system; - Providing digitized geo-referenced accident data, comprising statistical accident data, in particular in the form of at least one database, which can be processed by the computer system; —- Data collection, in particular by the computer system, of geo-referenced user input data related to potential danger areas; —- Data acquisition, in particular by the computer system, of geo-referenced sensor data related to the movement characteristics of road users, in particular based on acceleration measurements which can be generated or are generated by sensors carried by road users or vehicles; - Assigning accident data, user input data and sensor data to segments, corresponding to georeferencing associated with these data, by the computer system, so that segments of the map-like image have associated recognition data, each comprising accident data, sensor data and / or user input data; - whereby the computer system carries out an evaluation of these recognition data at least for segments with recognition data, comprising at least one frequency determination and/or a data comparison, and whereby the computer system identifies a georeferenced segment with recognition data to which no accident data is assigned as a potential danger spot for the purpose of early detection, - if user input data and sensor data have been assigned to this segment with a predetermined frequency; and/or - if a data comparison of user input data and/or sensor data for this segment shows a predefined degree of feature agreement or feature correlation; and - wherein the computer system identifies segments identified as danger spots in the map-like image and preferably assigns them a computationally determined danger score. 2. Computer-aided method for determining a danger score of a georeferenced, structural danger point in a digital traffic route network map, in particular a segment identified as a danger point according to the method according to claim 1, by at least one computer system, the method comprising at least: - data collection of georeferenced user input data related to potential danger points; - Assigning provided geo-referenced accident data, collected georeferenced user input data and collected georeferenced sensor data into segments, corresponding to georeferencing associated with these data, by the computer system, so that segments of the map-like image have associated recognition data, each comprising accident data, sensor data and / or user input data; - computationally determining a danger score for segments of the map-like image with associated detection data by evaluating this detection data, - the danger score of a georeferenced segment with detection data to which no accident data is assigned is determined, in particular validated or relatively increased, as a function of assigned user input data; and/or - wherein the danger score is determined depending on a weighting of the detection data, each with a different weight for accident data, sensor data and user input data, in particular with a higher weight of accident data compared to user input data and/or a higher weight of user input data compared to sensor data; - Assigning the determined danger scores to the segments with assigned detection data. 3. The method according to claim 1 or 2, characterized in that the movement-related sensor data include a type feature which indicates which type of road user the sensor data comes from and enables a distinction between at least two different types of road users, in particular motor vehicles and two-wheelers; and/or at least for segments with recognition data for these recognition data, in particular for sensor data with type characteristics, in each case a computer pattern recognition, in particular AI-supported pattern recognition, is carried out and a classification into critical events takes place, in particular by the computer system, and the data comparison of user input data and /or sensor data with corresponding data to classify critical events, in particular the computer system identifying a georeferenced segment with detection data to which no accident data is assigned as a potential danger spot for the purpose of early detection if a data comparison of the sensor data for this segment shows sufficient feature correspondence or feature correlation with sensor data of a critical event. 4. The method according to claim 1, 2 or 3, characterized in that segments identified as danger spots are each assigned a data object which comprises at least one computationally determined danger score, the data object further comprising: - at least one relevance indication for which or which of Several different types of road users, in particular pedestrians, cyclists or motor vehicle drivers, are threatened with danger, the relevance information preferably being determined from the associated recognition data; and/or - at least one danger type indication as to which traffic behavior, which situation and/or between which road users there is danger at the danger zone, the danger type indication preferably being determined from the associated detection data; and/or - at least one contextual indication as to which environmental influences or traffic situations pose a threat of danger at the danger point, the contextual indication preferably being determined from the associated recognition data and/or from additional traffic route metadata, in particular comprising information regarding weather dependencies, speed guidelines, time-dependent traffic volume or like.. 5. The method according to one of claims 1 to 4, characterized in that the data collection of geo-referenced user input data related to potential danger spots takes place continuously and / or sporadically via an Internet interface of the computer system, which is connected to at least one server, in particular web server, which provides a user interface for user input data. 6. Method according to one of claims 1 to 5, characterized in that dividing the map-like image into a plurality of georeferenced segments includes variable dividing into segments of different geographical lengths, in particular depending on the traffic infrastructure and/or depending on associated recognition data and/or or depending on identified danger spots, so that there is sufficient local discrimination of danger spots; and/or assigning accident data, sensor data and user input data to individual segments includes determining a geographical distance between the detection data from their georeferencing and the assignment is based on relative distances. 7. The method according to one of claims 1 to 6, characterized in that for data acquisition, sensor data is acquired based on acceleration measured values from different sources, in particular sensor data of a first type, which consist of sensor values from the vehicle equipment, in particular vehicle sensor systems or telematics equipment, of vehicles, in particular of motor vehicles, as well as at least sensor data of a second type, which are generated from sensor values from acceleration sensors in portable mobile devices carried by road users, in particular smartphones; and/or the sensor data are preferably provided with time stamps, the time stamps being used in the data comparison with regard to a predefined degree of feature agreement or feature correlation, in particular as to whether the time stamps are spaced apart in time within a predefined time range or correlate based on the time of day. 8. The method according to one of claims 1 to 7, in particular according to claim 7, characterized in that georeferenced sensor data are provided by an application-specific smartphone application for detecting potential danger spots, wherein - the smartphone application is preferably set up to detect potentially safety-critical events detect, in particular on the basis of exceeding acceleration threshold values, and provide geo-referenced sensor data only for potentially safety-critical events, the sensor data provided preferably containing information about movement behavior within a time window around the safety-critical event; and/or - the smartphone application provides an interface for georeferenced user input data related to potential danger spots. 9. The method according to claim 8, characterized in that - when a potentially safety-critical event is detected, the smartphone application assigns corresponding sensor data to the associated GPS coordinates for georeferencing, and preferably at least one time stamp, and transmits them to the computer system with the sensor data; - the smartphone application, in particular when detecting a potentially safety-critical event and / or when detecting local proximity to a segment identified as a potential danger point, asks the user to enter user input data, comprising at least one confirmation regarding the existence of a danger point, preferably further for input a relevance information regarding affected road user types and/or a hazard type information and/or a context information and the user input is transmitted to the computer system as user input data with the sensor data; and/or - the smartphone application provides relevance information regarding affected road user types based on a movement profile determined by means of an integrated acceleration sensor for a potentially safety-critical event and transmits it to the computer system with the sensor data. 10. The method according to any one of the preceding claims, characterized by computationally determining a danger score for an identified danger spot by calculating, the calculation comprising a weighting of georeferenced assigned sensor data using predefined weighting factors, and preferably at least some georeferenced sensor data of a first type depending on a predefined degree of agreement with features of georeferenced assigned user input data and / or further georeferenced assigned sensor data of a second type are weighted; and/or computationally determining a danger score for an identified danger point by calculating, the calculation comprising processing by means of an AI, in particular using a neural network and/or pattern recognition by the at least one computer system, taking into account event-related characteristics of the infrastructure data. 11. The method according to one of the preceding claims, in particular according to claim 4, characterized by computationally determining at least one hazard type information for each identified hazard location, the hazard type information consisting of information from infrastructure data and / or sensor data and / or accident reports and / or user input data assigned to the segment, preferably is determined from a combination of information from at least two different of these data. 12. Method according to one of the preceding claims, characterized by assigning a danger level from a number of discrete ones Danger level for each danger score, with each danger level being assigned a predefined value range of the danger score; and/or computationally determining multiple danger scores for an identified danger spot depending on the respective relevance information as to which of several different types of road users, in particular pedestrians, cyclists or motor vehicle drivers, are at risk of danger and assigning a respective danger score to different types of road users. 13. The method according to any one of the preceding claims, characterized by gradually updating the data included in the digital mapping of traffic route networks, in particular the danger scores, when a predetermined amount of further sensor data has been recorded and / or a predetermined amount of further user input data has been collected. 14. The method according to one of the preceding claims, in particular according to claim 13, characterized by adapted data acquisition of geo-referenced sensor data, related to the movement behavior of a road user, in particular with increased sensitivity or intensity, upon detection of the road user's local proximity to one that has already been identified as a potential danger spot segments. 15. Computer system, in particular cloud computer system, comprising a map-like digital traffic route network image, which has segments with a danger score identified as danger spots according to a method according to one of the preceding claims 1 to 13. 16. A method for generating danger zone-related warning messages to road users, with terminal devices communicating with the computer system according to claim 15, in particular smartphones and / or navigation devices, upon detection of local proximity to a segment identified as a danger zone Issue a warning message, in particular a warning message comprising a relevance information and/or a hazard type information and/or a context information. 17. Use of the computer system according to claim 15 for determining safe route options in a navigation system, in particular in addition to fast and / or ecological route options determined by the navigation system. 18. Use of the computer system according to claim 15 for communication with autonomous or semi-autonomous driving vehicles in order to influence the autonomous or semi-autonomous driving behavior depending on danger spots, in particular taking into account relevance and/or danger type and/or context information and/or for the purpose of influencing route selection. 19. Use of the computer system according to claim 15 to provide a data set for training a machine learning model to predict potential danger spots, the machine learning model in particular for optimizing the map-like digital traffic route network mapping and / or in a simulation for traffic planning is being used.