TWI674560B - Behavior analysis system and method of vehicle - Google Patents

Abstract

A vehicle behavior analysis system and method. This behavior analysis method includes the following steps. Detect eTag information of vehicles passing by the lane. Capture an image that includes the license plates of passing vehicles. Obtain identification information of vehicles detected in those lanes in several time intervals. For each time interval, the detected eTag information on the same lane is matched with the detected license plate to generate multiple sets of matching combinations. Each pairing combination records all detected eTag information and license plates on the same lane in a certain time interval. The pairing combinations in different time intervals are compared to determine whether the identification information of each vehicle conforms to the identification change behavior, and the identification change behavior is that the pairing combinations in different time intervals do not match. Thereby, the functions of automatic vehicle identification and behavior analysis can be provided.

Description

Vehicle behavior analysis system and method

The invention relates to a vehicle identification and analysis technology, and in particular to a vehicle behavior analysis system and method.

After the cancellation of national tolls, the mounting rate of eTags (or Radio Frequency Identification (RFID)) on vehicles is quite high. According to statistics from the Ministry of Communications, more than 80% of vehicles have eTags installed. On the other hand, because the law requires vehicles to carry license plates before they are allowed to go on the road, generally normal users will install license plates on their vehicles. However, there are still some users who fail to comply with traffic regulations to publish legal license plates or change license plates. License plates and eTags are important identification information used to identify vehicles. However, eTags and license plates are personal data and are protected by personal information laws. Unless necessary reasons cannot be used to directly query the user's license plate through eTag. In addition, the current search methods for some abnormal behaviors (such as changing license plates by themselves, forging license plates, etc.) are usually first discovered by traffic police as abnormal behaviors of passersby (such as speeding, modification, speeding, etc.) ),again By connecting to the database of the supervisory authority through the handheld device, you can know whether the query vehicle has abnormal behavior. That is to say, the existing query methods of vehicle identification information are all manual, and there is no automated mechanism.

In view of this, the present invention provides a vehicle behavior analysis system and method, which provides an automatic dual identification mechanism and judges the identification change behavior based on the information obtained from the identification.

In the vehicle behavior analysis system of the present invention, the identification information of each vehicle includes an eTag and a license plate. The behavior analysis system includes a front-end identification system and a back-end identification system. The front-end identification system shoots one or more lanes on a certain road section, and includes an eTag reader and an image capture unit. The eTag reader detects the eTag information of passing vehicles through the connected eTag antenna, and each eTag antenna corresponds to a lane. Each image capturing unit corresponds to a lane and captures an image including a license plate of a driving vehicle. The back-end identification system obtains identification information of vehicles detected on those lanes in several time intervals. For each time interval, the back-end identification system matches the eTag information detected on the same lane with the license plate detected on the same lane to generate multiple sets of matching combinations. Each pairing combination records all detected eTag information and license plates on the same lane in a certain time interval. The back-end identification system compares pairing combinations in different time intervals to determine whether the identification information of each vehicle conforms to the identification change behavior, and the identification change behavior is that the pairing combinations in different time intervals do not match.

According to the vehicle behavior analysis method of the present invention, the identification information of each vehicle includes eTag and license plate. The behavior analysis method includes the following steps. Detect eTag information of vehicles passing by one or more lanes. Capture an image that includes the license plates of passing vehicles. Obtain identification information of vehicles detected in those lanes in several time intervals. For each time interval, the eTag information detected on the same lane is matched with the license plate detected on the same lane to generate multiple sets of matching combinations. Each pairing combination records all detected eTag information and license plates on the same lane in a certain time interval. The pairing combinations in different time intervals are compared to determine whether the identification information of each vehicle conforms to the identification change behavior, and the identification change behavior is that the pairing combinations in different time intervals do not match.

Based on the above, the vehicle behavior analysis system and method of the embodiments of the present invention are based on the pairing of eTag and license plate identification, and generate a set of learning mechanisms through a large number of matching combinations of license plates and eTags, and automatically determine the user's behavior based on this. , So that the behavior judgment results can be used as a reference for supervision stations, transportation brigades, or criminal police teams.

In order to make the above features and advantages of the present invention more comprehensible, embodiments are hereinafter described in detail with reference to the accompanying drawings.

1‧‧‧Behavior Analysis System

10‧‧‧ Front-end identification system

11, 51‧‧‧communication transceiver

13, 13a, 13b‧‧‧eTag antenna

14‧‧‧eTag Reader

15‧‧‧Image recognition device

151, 151a, 151b‧‧‧Image Acquisition Unit

153‧‧‧License plate recognition module

19, 59‧‧‧ processor

L1, L2‧‧‧ lane

50‧‧‧back-end identification system

57‧‧‧Storage

571‧‧‧License plate recognition module

573‧‧‧Behavior Analysis Module

575‧‧‧trust database

577‧‧‧ Training Database

S310 ~ S390, S410 ~ S490‧‧‧step

FIG. 1 is a component block diagram of a behavior analysis system according to an embodiment of the present invention.

FIG. 2 is a schematic diagram of a front-end identification system according to an embodiment of the present invention.

3 is a flowchart of a behavior analysis method-training phase according to an embodiment of the present invention.

4 is a flow of a behavior analysis method-comparison phase according to an embodiment of the present invention Process map.

FIG. 1 is a block diagram of components of a behavior analysis system 1 according to an embodiment of the present invention. Referring to FIG. 1, the behavior analysis system 1 includes a front-end recognition system 10 and a back-end recognition system 50.

The front-end identification system 10 includes, but is not limited to, a communication transceiver 11, one or more eTag antennas 13, one or more eTag readers 14, an image recognition device 15, and a processor 19.

The communication transceiver 11 can support third generation (3G), fourth generation (4G) or later generation mobile communication, Ethernet or fiber modem or other network access devices, so as to interact with the outside world. communication.

The eTag (or RFID) reader 14 detects the eTag information of the passing vehicle through the connected eTag antenna 13, and each eTag antenna 13 corresponds to a lane.

The image recognition device 15 includes one or more image capturing units 151 and a license plate recognition module 153. Each image capturing unit 151 includes components such as a lens and an image sensor, and is a device such as a camera or a video camera, and is used to capture an image of a lane. The license plate recognition module 153 may be an image recognition chip, a processor, or a circuit, and is used to recognize the license plate in the image.

The processor 19 is coupled to the communication transceiver 11, the eTag reader 14, and the image recognition device 15. The processor 19 may be a central processing unit (CPU), a microcontroller, a programmable controller, and a special application integrated circuit ( Application Specific Integrated Circuit, ASIC) or other similar components or a combination of the above.

Please refer to FIG. 2, which is a schematic diagram of a front-end identification system 1 according to an embodiment of the present invention. The front-end identification system 1 is set on a certain road section. This section includes lanes L1, L2, eTag antenna 13a and image capture unit 151a correspond to lane L1, and eTag antenna 13b and image capture unit 151b correspond to lane L2. That is to say, the vehicle identification information (including eTag information and license plate) passing the lane L1 can be detected by the eTag antenna 13a and the image capturing unit 151a, and the vehicle identification information passing the lane L2 can be captured by the eTag antenna 13b and the image Unit 151b detected. It should be noted that the positions and numbers of the components and lanes shown in FIG. 2 can be adjusted according to requirements, which is not limited in the embodiment of the present invention.

It is worth noting that the reading range of the eTag antennas 13a and 13b is different from the image capturing unit 151 and usually forms a fan-shaped reading range. In most multi-lane application scenarios, the eTag antenna 13 may read the eTags of adjacent lanes or front and rear vehicles, and the pairing error rate may increase significantly. The embodiments of the present invention will improve the success rate of pairing through a large amount of data analysis, which will be detailed in the subsequent embodiments.

The back-end identification system 50 includes, but is not limited to, a communication transceiver 51, a storage 57 and a processor 59. The backend identification system 50 may be a desktop computer, a server, or a workstation.

The communication transceiver 51 may support 3G, 4G or later generation mobile communication, Ethernet or fiber optic modem or other network access devices, so as to communicate with the outside world.

The memory 57 may be any type of fixed or removable random access memory RAM (Radom Access Memory, RAM), Read Only Memory (ROM), flash memory (flash memory), traditional hard disk drive (HDD), solid-state drive (solid-state drive, SSD) or similar components and used to record code, software modules (for example, license plate recognition module 571, behavior analysis module 573, etc.), trust database 575, training database 577, images, pairing combinations and other data or Archives, the details of which will be detailed in the following embodiments.

The processor 59 is coupled to the communication transceiver 51 and the memory 57. The processor 59 may be a CPU, a microcontroller, a programmable controller, an ASIC, or other similar components or a combination of the foregoing components.

In order to facilitate the understanding of the operation flow of the embodiments of the present invention, a number of embodiments will be described in detail below for the collection, training, and analysis processes for vehicle recognition in the embodiments of the present invention. Hereinafter, the method described in the embodiment of the present invention will be described with various devices, components, and modules in the behavior analysis system 1. Each process of the method can be adjusted according to the implementation situation, and is not limited to this.

3 is a flowchart of a behavior analysis method-training phase according to an embodiment of the present invention. Referring to FIG. 3, the eTag reader 14 of the front-end identification system 1 detects the eTag information of vehicles passing by one or more lanes through the eTag antenna 13 corresponding to each lane, and each image capturing unit 151 captures the vehicles including passing vehicles. To obtain the identification information of the vehicles detected in those lanes within several time intervals (for example, 0.5, 1, 2 seconds, etc.), thereby collecting data on the identification information of the vehicle (step S310). These time intervals can be set to be the same or adjusted as needed. In other words, the front-end identification system 1 can obtain a certain path every time interval set. ETag information and license plate image detected by the segment.

Next, the license plate recognition module 153 in the front-end recognition system 1 recognizes the license plate image (step S330). The license plate recognition module 153 uses, for example, a hardware recognition mechanism to recognize the license plate (ie, license plate number) in the image, that is, the license plate recognition module 153 is a special chip or processing circuit designed for license plate recognition.

It is worth noting that the license plate recognition module 153 is the first built-in hardware license plate recognition, which is characterized by faster recognition speed. However, due to factors such as weather, lighting, and missing license plates, it will inevitably lead to misunderstanding. Therefore, when the first license plate recognition module 153 detects the vehicle, the first license plate recognition will be performed first to obtain the front-end license plate recognition result. The processor 19 transmits the front-end license plate recognition result, the image including the license plate (that is, the image captured by the image capture unit 151), and the detected eTag information to the back-end recognition system 50 through the communication transceiver 11.

The processor 59 of the back-end recognition system 50 can obtain the identification information of the vehicle detected by the front-end recognition system 10 on the lane in several time intervals through the communication transceiver 51, and the license plate recognition module 571 performs the second lane on the license plate image. License plate recognition (step S350). The license plate recognition module 571 can be detected by software (for example, AdaBoost algorithm, Principal Component Analysis (PCA), Independent Component Analysis (ICA), artificial intelligence (AI) classifier, etc.) Filter irrational data (for example, the license plate is cut, the picture is obviously indeterminate, etc.), and use it to obtain the back-end license plate recognition result. The license plate recognition module 571 can also increase the accuracy of the judgment through software self-learning (for example, machine learning, etc.).

The license plate recognition module 571 compares the recognition result of the back-end license plate with the recognition result of the front-end license plate corresponding to the same license plate image to determine whether the two results match (step S370). However, when both the license plate recognition steps S330 and S350 are the same license plate (ie, the license plate number is the same), the license plate recognition module 571 will determine that the front-end and rear-end license plate recognition results are correct and reasonable.

For these license plate recognition results that are judged to be correct and reasonable (matching), the processor 59 performs cross-comparison with the eTag information (step S390) to form pairing combinations, thereby storing these pairing combinations in the training database 577. In this embodiment, for each time interval, the processor 59 pairs the eTag information detected on the same lane with the license plate detected on the same lane to generate multiple sets of pairing combinations. For example, in the case of multiple lanes, the processor 59 needs to first correct the time of the license plate recognition modules 153, 571 and the eTag reader 14 and confirm that the time is the same (that is, the time interval of the eTag reader 14 reading the eTag information). And the corresponding time interval of the image recognized by the license plate recognition module 153, 571), and then start pairing. The front-end and back-end identification systems 10, 50 first collect identification data for a period of time as a basis for matching, and this phase is called a "training stage". The processor 59 matches the license plate recognition result of the same lane in the same time interval (for example, within the same second) with the eTag information detected in the same lane in the same time interval (for example, within the same second). That is, each pairing combination records all detected eTag information and license plates on the same lane in a certain time interval. And because the reading range of the eTag antenna 13 may involve more than two lanes, a single license plate may correspond to multiple eTag information. The processor 59 will write all possible pairing combinations into the `` training table '' and record the matching at the same time The right time interval is recorded in the training database 577.

4 is a flowchart of a behavior analysis method-comparison phase according to an embodiment of the present invention. Relevant parameters (e.g., starting time, starting position, the multi-lane road, until the point in time, until the point position, etc.) of the vehicle behavior processor 59 initializes, to set the width of the active state observation time (△ t _s) (i.e., statistical time Pitch, used to select pairing combinations for some time intervals), and initialize the timing number n to 0 (step S410). The processor 59 then counts the pairing combinations in the training database 577 at a specific statistical time (eg, one month, three months, or six months, etc.). When the amount of collected data is sufficient (e.g., one month, or more than six months of vehicle matching combinations), for example, if every 30 minutes is set as a time period (that is, a statistical time interval), the processor 59 may The day is divided into 48 periods (ie, 48 statistical time intervals). The processor 59 will determine whether to write into the trusted database 575 through the calculation formula (1): | Di × Wi | <T ... (1) Di is the time difference between the first and second time intervals, and this first time interval (t = t _n ) corresponds to the most recent (new) pairing combination, and the second time interval corresponds to the last same pairing combination before the first time interval; Wi is weight; T is threshold. In other words, the processor 59 performs a weighting operation on the time difference Di (step S450), and determines whether the weighted operation result of the time difference is smaller than the threshold value T (step S470). If the result of the weighting operation is less than the threshold value T, the processor 59 inputs (records) the pairing combination corresponding to the time interval to the trust database 575 (step S480). Conversely, if this operation does not result weighted less than the threshold T, the statistics processor 59 will adjust the time interval (time width observed △ t _s) (returning to step S410). It should be noted that the threshold T can be adjusted according to the parameters set in practice.

Next, the behavior analysis module 573 performs an automatic comparison process on the pairing combinations recorded in the trust database 575 (step S). In this embodiment, the behavior analysis module 573 can (for example, through SQL program syntax) compare matching combinations of different time intervals within a specific observation interval width to determine whether the identification information of each vehicle conforms to the identification change behavior ( Step S490).

The identification change behavior is that the pairing combinations in different time intervals do not match. For example, suppose time interval t2 is after time interval t1 (that is, time interval t2 is later than time interval t1), the license plate is P _i (i is a positive integer), and eTag information is E _j (j is a positive integer). If there is one eTag information corresponding to two (or more) license plates in several matching combinations in the statistical time interval, the time corresponding to the eTag information and the license plate can be observed through time sorting. If the eTag information E ₁ in the pairing combination counted before the time interval t1 corresponds to the license plate P ₁ , but after this time interval t ₁ corresponds to a different license plate P ₂ , the behavior analysis module 573 judges that the identification change behavior is a license plate replacement behavior. If the eTag information E ₁ in the pairing combination counted before the time interval t1 corresponds to the license plate P ₁ and after the time interval t1 corresponds to the license plate P ₂ , but after the time interval t ₂ corresponds to the license plate P ₁ , the behavior analysis module 573 judges this Recognizing the change behavior is to change the license plate back and forth.

If there are two (or more) eTag information corresponding to the same license plate in several matching combinations, you can still observe all the matching combinations by time sorting. If the pairing combination of time interval t1 is the license plate P ₃ corresponding to the eTag group (including at least two different eTag information), and the pairing combination of time interval t2 is still the license plate P ₃ corresponding to the same eTag group, the behavior analysis module 573 determines that this identification change behavior is the installation of plural eTag behaviors. And if the license plate P _{3 in the} pairing combination counted before the time interval t1 corresponds to the eTag information E ₂ , but the license plate P ₃ in the pairing combination counted after the time interval t ₁ corresponds to the eTag information E ₃ , the behavior analysis module 573 judges the identification change The behavior is to replace the eTag behavior.

It is worth noting that the aforementioned comparison process will enter a loop operation, and the behavior analysis module 573 will receive a pairing combination of eTag information from the same time interval and the license plate and distinguish it by lanes, and according to the previous training stage (training stage) Record the corresponding historical pairing combination in the trust database 577, and after updating the latest data, determine whether there is a special identification change between the license plate and eTag information of the currently passed vehicle.

In summary, the embodiment of the present invention installs a front-end identification system on one or more lanes on the main road or any road section to be detected, and does not need to know the eTag information corresponding to the official data of the license plate. An automated way to pair eTag information with a license plate. Next, the embodiment of the present invention collects historical records of driving behaviors in the past, and then uses algorithms and thresholds to determine all possible combinations of eTag information and license plates. These matching combinations can be further used as the basis for behavior analysis and feedback from related personnel, even for criminal cases.

The embodiment of the present invention includes the following features and effects: In the embodiment of the present invention, eTag information is matched with a license plate without obtaining the true correspondence between the eTag information and the license plate recorded in the official database, and the reasonable matching combination is automatically determined through historical records.

The embodiment of the present invention uses an automatic filtering method: considering that there may be several vehicles passing through the lane at the same time, the corresponding relationship between the eTag information detected by the front end and the vehicle is It is not necessarily correct, and it is necessary to confirm a reasonable pairing combination through a larger number of pairing combinations, so as to avoid using a single threshold to cause wrong pairing judgment.

In the embodiment of the present invention, individual statistics are classified and classified through different pairing combinations to select statistical data suitable for subsequent behavior analysis.

Although the present invention has been disclosed as above with the examples, it is not intended to limit the present invention. Any person with ordinary knowledge in the technical field can make some modifications and retouching without departing from the spirit and scope of the present invention. The protection scope of the present invention shall be determined by the scope of the attached patent application.

Claims (10)

A vehicle behavior analysis system, in which the identification information of each vehicle includes an eTag and a license plate, and the behavior analysis system includes: a front-end identification system, which is located in at least one lane of a road section, and includes: an eTag reader, A plurality of connected eTag antennas detect eTag information of passing vehicles, wherein each of the eTag antennas corresponds to a lane; and a plurality of image capture units, wherein each of the image capture units corresponds to a lane, and captures Taking a plurality of images including the license plates of passing vehicles; and a license plate recognition module to recognize the license plates in the images; a back-end recognition system to obtain the identification of the vehicles detected on the at least one lane in multiple time intervals Information; for each time interval, the eTag information detected on the same lane is paired with the license plate detected on the same lane to generate multiple matching combinations, each of which is recorded in the matching combination -All detected eTag information and license plates on the same lane in the time interval; and pairing combinations that compare in different time intervals to determine each of the vehicles Whether the identification information of the vehicle conforms to an identification change behavior, wherein the identification change behavior is that the pairing combinations in different time intervals do not match. According to the vehicle behavior analysis system described in item 1 of the patent application scope, wherein the time intervals include a first time interval and a second time interval, and the back-end identification system, Weight the time difference between the first time interval and the second time interval, wherein the first time interval corresponds to the latest pairing combination, and the second time interval is the same as the previous time before the first time interval Pairing combination; judging whether the weighted operation result of the time difference is less than a threshold value; and if the weighted operation result is less than the threshold value, inputting the pairing combination corresponding to the second time interval into a trust database, from the trust database The data in the judgment of the identification change behavior. For example, the vehicle behavior analysis system described in item 2 of the scope of patent application, wherein if the weighted calculation result is not less than the threshold value, the back-end recognition system adjusts a statistical time interval, wherein the statistical time interval is used to select some of these Paired combinations of time intervals. According to the vehicle behavior analysis system described in item 1 of the patent application scope, the time intervals include a third time interval and a fourth time interval, and the fourth time interval is after the third time interval, and if the The first eTag information in the pairing combination counted before the third time interval corresponds to a first license plate, but after the third time interval corresponds to a different second license plate, the back-end recognition system determines that the recognition change behavior is a license plate change Behavior; if the first eTag information in the statistical pairing combination before the third time interval corresponds to the first license plate, and after the third time interval corresponds to the second license plate, but after the fourth time interval, it corresponds to the first license plate A license plate, the back-end recognition system judges that the recognition change behavior is a license plate replacement behavior; if the pairing combination in the third time interval is a third license plate corresponding to an eTag Group, and the matching combination of the fourth time interval is still that the third license plate corresponds to the eTag group, the back-end identification system determines that the identification change behavior is the installation of multiple eTag behaviors, where the eTag group includes multiple Different eTag information; and if the third license plate in the pairing combination counted before the third time interval corresponds to a second eTag information, but the third license plate in the pairing combination counted after the third time corresponds to a third eTag information , The backend identification system determines that the identification change behavior is an eTag replacement behavior. According to the vehicle behavior analysis system described in the first item of the scope of patent application, the back-end recognition system obtains the images obtained by the image capturing units and obtains a back-end license plate recognition result to determine the back-end license plate recognition. The result is consistent with the recognition result of the license plate in the passing vehicle by the front-end recognition system, and the matching license plate is matched with corresponding eTag information. A vehicle behavior analysis method, in which identification information of each vehicle includes an eTag and a license plate, and the behavior analysis method includes: detecting eTag information of passing vehicles in at least one lane; capturing a plurality of images including the license plates of passing vehicles; obtaining Identification information of vehicles detected on the at least one lane in multiple time intervals; for each of the time intervals, the eTag information detected on the same lane is compared with the detected eTag information on the same lane Vehicle license plate pairing to generate multiple matching combinations, each of which records a detected eTag information and license plate on the same lane in the same time interval; and The pairing combinations of the different time intervals are compared to determine whether the identification information of each of the vehicles conforms to an identification change behavior, wherein the identification change behavior is that the pairing combinations of the different time intervals do not match. According to the vehicle behavior analysis method described in item 6 of the scope of the patent application, wherein the time intervals include a first time interval and a second time interval, and the step of comparing the pairing combinations of the different time intervals includes: The time difference between the first time interval and the second time interval is weighted, wherein the first time interval corresponds to the latest pairing combination, and the second time interval corresponds to the previous pairing combination before the first time interval. ; Determine whether the weighted operation result of the time difference is less than a threshold value; and if the weighted operation result is less than the threshold value, then input the pairing combination corresponding to the second time interval into a trust database, and from the trust database The data determines the identification change behavior. The behavior analysis method for a vehicle as described in item 7 of the patent application scope, wherein after the step of determining whether a weighted operation result of the time difference is less than the threshold value, the method further includes: if the weighted operation result is not less than the threshold value, the back end The identification system adjusts a statistical time interval, wherein the statistical time interval is used to select a pair of combinations of the time intervals. The behavior analysis method for a vehicle according to item 6 of the scope of patent application, wherein the time intervals include a third time interval and a fourth time interval, and the fourth time interval The time interval is after the third time interval, and the step of comparing pairing combinations in different time intervals includes: if a first eTag information in a pairing combination counted before the third time interval corresponds to a first license plate, but After the third time interval corresponds to a different second license plate, it is judged that the identification change behavior is a license plate replacement behavior; if the first eTag information in the statistical matching combination before the third time interval corresponds to the first license plate, and the After the third time interval corresponds to the second license plate, but after the fourth time interval corresponds to the first license plate, it is judged that the identification change behavior is a back and forth license plate behavior; if the pairing combination of the third time interval is a third The license plate corresponds to an eTag group, and the pairing combination in the fourth time interval is still the third license plate corresponds to the eTag group, then it is determined that the identification change behavior is the installation of multiple eTag behaviors, where the eTag group includes multiple Different eTag information; and if the third license plate in the pairing combination counted before the third time interval corresponds to a second eTag information, but after the third time The third license statistics pairing combinations correspond to a third eTag information, it is determined that the identification of behavior change is the replacement of eTag behavior. According to the method for analyzing the behavior of a vehicle as described in the sixth item of the patent application, before the step of comparing the pairing combinations of the different time intervals, the method further includes: obtaining the images and obtaining a back-end license plate recognition result; judging Whether the back-end license plate recognition result matches the previous recognition result of the license plate in the passing vehicle; and Match the matching license plate with the corresponding eTag information.