A method, apparatus, equipment and medium for filtering license plate recognition errors

Abstract

This invention provides a method, apparatus, device, and medium for filtering license plate recognition errors. The method includes acquiring vehicle passage record data and vehicle data stored in the vehicle management office; comparing the vehicle passage record data with the vehicle data stored in the vehicle management office to obtain suspected fake license plate data initially identified as fake, and storing the suspected fake license plate data in a first index; for the suspected fake license plate data initially identified as fake, querying all vehicle passage records within a preset time period based on the vehicle passage record data to obtain the vehicle's trajectory data; calculating the joint recognition rate of checkpoints based on the vehicle trajectory data, and filtering the suspected fake license plate data based on the joint recognition rate of checkpoints. The advantages of this invention are: by utilizing data from multiple checkpoints and vehicle trajectory information, and combining the joint recognition rate of checkpoints and the reasonable probability of trajectory data to filter the fake license plate data initially identified as fake, the reliability and accuracy of license plate recognition can be improved, and problems caused by incorrect recognition can be reduced.

Description

Technical Field

[0001] This invention relates to the field of intelligent transportation technology, and in particular to a method, apparatus, device, and medium for filtering license plate recognition errors. Background Technology

[0002] With the development of information technology and the continuous increase in the number of motor vehicles, various traffic violations and irregularities are also on the rise. Among these, the use of counterfeit license plates poses a significant threat. Counterfeit license plates refer to the use of license plate recognition technology to detect and prevent the use of forged, altered, or illegal license plates in areas such as traffic management, security monitoring, and public safety.

[0003] License plate recognition systems typically rely on technologies such as image processing, computer vision, and machine learning to automatically monitor and manage vehicles by recognizing their license plate numbers. Currently, counterfeit license plate analysis primarily compares license plate numbers identified by checkpoint records with those registered at vehicle management offices. If a license plate cannot be found in the office's records, it is recorded as a suspected counterfeit. License plate recognition systems can face various problems, leading to numerous suspected counterfeit license plate errors. This significantly increases the workload for manual verification. Therefore, filtering out erroneous license plate data from suspected counterfeit plates is crucial to greatly reduce manual judgment and improve efficiency. To reduce erroneous license plate data and improve recognition accuracy, existing technologies typically employ the following two methods:

[0004] The first approach is to improve the accuracy of fake license plate analysis. From the root cause, this means improving the accuracy of the license plate recognition system. However, in practical applications, license plate recognition systems may face various problems that lead to frequent misidentifications. For example, factors such as blurry images, insufficient lighting, reflections, and occlusions can all cause deviations in the quality of license plate images, thereby affecting the accuracy of license plate recognition.

[0005] The second method, proposed in publication CN107329977B, is a secondary screening method for fake license plates based on probability distribution. It determines the error probability of each character recognition based on the spatial probability distribution of vehicle flow, and comprehensively judges the probability of a license plate being fake based on the probability of the license plate conforming to the spatial distribution and the character recognition error probability. However, the effectiveness and accuracy of this method depend on the quality and completeness of the acquired checkpoint vehicle passage record data. Incomplete or noisy checkpoint vehicle passage record data will also affect the accuracy of license plate recognition. As can be seen from the above, existing technologies still suffer from poor filtering of license plate recognition error data, which in turn affects the accuracy of license plate recognition. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to provide a method, device, equipment and medium for filtering license plate recognition errors, so as to solve the problem that the existing technology still has the problem of poor filtering of license plate recognition error data when recognizing license plates, which will affect the accuracy of license plate recognition.

[0007] In a first aspect, the present invention provides a method for filtering license plate recognition errors, the method comprising the following steps:

[0008] The system obtains vehicle passage record data and vehicle data stored in the vehicle management office. It compares the vehicle passage record data with the vehicle data stored in the vehicle management office to obtain suspected fake license plate data that is initially identified as fake license plates, and stores the suspected fake license plate data in the first index.

[0009] For suspected fake license plate data that is initially identified as fake, the vehicle's trajectory data is obtained by querying all vehicle passage records within a preset time period based on the vehicle passage record data; the joint recognition rate of checkpoints is calculated based on the vehicle trajectory data, and the suspected fake license plate data is filtered based on the joint recognition rate of checkpoints.

[0010] Furthermore, the method also includes: calculating the probability of trajectory rationality based on vehicle trajectory data and location association configuration table, and filtering suspected fake license plate data based on the trajectory rationality probability; when the location association configuration table is initialized, it calculates the probability distribution of the uplink and downlink locations corresponding to each location by calculating the vehicle trajectory data over a period of time.

[0011] Furthermore, the acquisition of vehicle passage record data and vehicle data stored in the vehicle management office specifically includes:

[0012] Retrieve vehicle passage records from Kafka, extract the license plate numbers from the vehicle passage records, and determine whether the extracted license plate numbers match the set regular expression. If they match, keep the vehicle passage records; otherwise, discard the vehicle passage records.

[0013] Retrieve vehicle registration information stored in the vehicle management office. The vehicle registration information includes license plate number, license plate type and other information. Put the license plate number and license plate type into Redis as the key and the other information as the value.

[0014] The process of comparing vehicle passage record data with vehicle data stored in the vehicle management office to obtain suspected fake license plate data initially identified as fake, and storing the suspected fake license plate data in the first index, specifically includes:

[0015] Based on the obtained vehicle passage records, the system queries Redis using the license plate number and license plate type as keys to check if corresponding data exists. If corresponding data exists, the license plate is considered genuine; otherwise, it is considered fake. The vehicle passage records are then stored in the first index of Elasticsearch, thus obtaining a preliminary list of suspected fake license plate data. The fields of the first index include primary key, license plate number, license plate type, passage time, location ID, device ID, and region ID.

[0016] Furthermore, the step of querying all vehicle passage records within a preset time period based on vehicle passage record data to obtain the vehicle's trajectory data specifically includes:

[0017] Retrieve a list of suspected fake license plates from the first index; loop through the suspected fake license plate list and, using the license plate number and type, query the vehicle passage records within a first preset time period before the passage time as the upward passage trajectory and the vehicle passage records within a second preset time period after the passage time as the downward passage trajectory, thus obtaining the vehicle's trajectory data. The first preset time period and the second preset time period are equal. Simultaneously, if the number of points in both the upward and downward passage trajectories is zero, delete the suspected fake license plate from the suspected fake license plate list in the first index.

[0018] Furthermore, the step of calculating the joint recognition rate of checkpoints based on vehicle trajectory data and filtering suspected fake license plate data based on the joint recognition rate specifically includes:

[0019] When the total number of points for the upbound and downbound vehicle trajectories is greater than or equal to the preset value, the joint recognition rate of the checkpoint is calculated as Score1 = N / M, where M is a constant set according to the checkpoint coverage rate, and N is the total number of points for the upbound and downbound vehicle trajectories.

[0020] If the calculated joint recognition rate score1 is greater than or equal to the preset first threshold, the suspected fake license plate data is classified as high-probability fake license plate data and stored in the second index. The fields of the second index include primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition. If not, the suspected fake license plate data is deleted from the suspected fake license plate data list in the first index.

[0021] Furthermore, the step of calculating the probability of trajectory reasonableness based on vehicle trajectory data and location association configuration table, and filtering suspected fake license plate data based on the probability of trajectory reasonableness, specifically includes:

[0022] When the total number of points in the upbound and downbound vehicle trajectories is less than a preset value, based on the point ID and upbound vehicle trajectory in the suspected fake license plate data, an upbound point association table corresponding to the upbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained; simultaneously, based on the point ID and downbound vehicle trajectory in the suspected fake license plate data, a downbound point association table corresponding to the downbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained, wherein the association probability of each point in the upbound and downbound point association tables is queried from the point association configuration table;

[0023] The trajectory rationality probability Score2 is calculated as A / C1 + B / C2, where A represents the sum of the association probabilities of each point in the uplink point association table, C1 represents the number of points in the uplink point association table, B represents the association probability of each point in the downlink point association table, and C2 represents the number of points in the downlink point association table.

[0024] If the calculated probability score2 of the trajectory is greater than or equal to the preset second threshold, the suspected fake license plate data is classified as high-probability fake license plate data and stored in the second index. The fields of the second index include primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition. If not, the suspected fake license plate data is deleted from the suspected fake license plate data list of the first index.

[0025] Secondly, the present invention provides a license plate recognition error filtering device, including a preliminary comparison module and a filtering module;

[0026] The preliminary comparison module is used to obtain vehicle passage record data and vehicle data stored in the vehicle management office, compare the vehicle passage record data with the vehicle data stored in the vehicle management office, obtain suspected fake license plate data that is initially identified as fake license plates, and store the suspected fake license plate data in the first index.

[0027] The filtering module is used to query all vehicle passage records within a preset time period based on the vehicle passage record data for suspected fake license plate data that is initially identified as fake license plates, thereby obtaining the vehicle's trajectory data; calculate the joint recognition rate of checkpoints based on the vehicle's trajectory data, and filter the suspected fake license plate data based on the joint recognition rate of checkpoints.

[0028] Furthermore, the filtering module is also used to calculate the probability of trajectory rationality based on the vehicle trajectory data and the point association configuration table, and to filter suspected fake license plate data based on the trajectory rationality probability; the point association configuration table, during initialization, calculates the probability distribution of the uplink and downlink points corresponding to each point by calculating the vehicle trajectory data over a period of time.

[0029] Thirdly, the present invention provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described in the first aspect.

[0030] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in the first aspect.

[0031] One or more technical solutions provided in the embodiments of the present invention have at least the following technical effects or advantages:

[0032] By utilizing data from multiple checkpoints and vehicle trajectory information, and combining the joint checkpoint recognition rate and trajectory reasonable probability to filter out fake license plate data initially identified as fake, the reliability and accuracy of license plate recognition can be improved, reducing problems caused by false recognition. Through trajectory data analysis, abnormal trajectories across checkpoints can be detected, further improving the screening effect for fake license plates. At the same time, it can handle license plate recognition errors caused by specific scenarios, vehicle movement states, or shooting conditions, enhancing the adaptability and robustness of the method.

[0033] The above description is merely an overview of the technical solution of the present invention. In order to better understand the technical means of the present invention and to implement it in accordance with the contents of the specification, and in order to make the above and other objects, features and advantages of the present invention more apparent and understandable, specific embodiments of the present invention are described below. Attached Figure Description

[0034] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0035] Figure 1 This is an execution flowchart of a license plate recognition error filtering method according to Embodiment 1 of the present invention;

[0036] Figure 2 This is a schematic diagram of the structure of a license plate recognition error filtering device according to Embodiment 2 of the present invention;

[0037] Figure 3 This is a schematic diagram of the electronic device in Embodiment 3 of the present invention;

[0038] Figure 4 This is a schematic diagram of the structure of the medium in Embodiment 4 of the present invention. Detailed Implementation

[0039] To better understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0040] Example 1

[0041] This embodiment provides a method for filtering license plate recognition errors, such as... Figure 1 As shown, the method includes the following steps:

[0042] Step S1: Obtain vehicle passage record data and vehicle data stored in the vehicle management office. Compare the vehicle passage record data with the vehicle data stored in the vehicle management office to obtain suspected fake license plate data that is initially identified as fake license plates, and store the suspected fake license plate data in the first index. Since the vehicle passage record data is collected and recorded by the license plate recognition system set up at the checkpoint, and the license plate recognition system may face various problems in actual use (such as image blurring, insufficient lighting, reflection, occlusion, etc.), it is difficult to avoid license plate recognition errors. Therefore, it is necessary to further filter the suspected fake license plate data that is initially identified as fake license plates.

[0043] Step S2: For suspected fake license plate data that is initially identified as fake, query all vehicle passage records within a preset time period based on the vehicle passage record data to obtain the vehicle's trajectory data; calculate the joint recognition rate of the checkpoints based on the vehicle's trajectory data, and filter the suspected fake license plate data based on the joint recognition rate of the checkpoints, thereby improving the reliability and accuracy of license plate recognition.

[0044] In this invention, the method further includes: calculating the probability of trajectory rationality based on vehicle trajectory data and a point-of-origin association configuration table, and filtering suspected fake license plate data based on the trajectory rationality probability to further improve the reliability and accuracy of license plate recognition; the point-of-origin association configuration table, during initialization, calculates the probability distribution of the uplink and downlink points corresponding to each point by calculating vehicle trajectory data over a period of time. The point-of-origin association configuration table is as follows:

[0045] Point ID Related point ID Related point types Association probability 11 22 Up 0.5 11 33 Up 0.3 11 44 Downward 0.7 11 none Up 0.2 11 none Downward 0.2 …

[0046] In a specific implementation of this invention, the point-to-point configuration table can be initialized by calculating the probability distribution of the uplink and downlink points corresponding to each point by calculating the vehicle trajectory data within a month. After initialization, it can also be manually adjusted. Specifically, when initializing the location association configuration table, vehicle passage trajectory data with license plates within the past month can be stored in the database. The SQL window functions `lag()` and `lead()` are then used to calculate the number of associations between each checkpoint and its downstream checkpoints, and the number of associations between each checkpoint and its upstream checkpoints within the past month, generating two intermediate result tables. These two intermediate result tables are then used to calculate the location association configuration table. For example, for checkpoint 1, which has checkpoints 2 and 3 downstream and checkpoint 4 upstream, the specific calculation involves first adding the association counts between checkpoint 2 and checkpoint 1, checkpoint 3 and checkpoint 1, and checkpoint 4 and checkpoint 1 to obtain the total association count for checkpoint 1. Then, the association count between checkpoint 2 and checkpoint 1 is divided by the total association count for checkpoint 1 to obtain the association probability between checkpoint 2 and checkpoint 1.

[0047] In some embodiments of the present invention, the acquisition of vehicle passage record data and vehicle data stored in the vehicle management office specifically includes:

[0048] Vehicle passage record data is retrieved from Kafka. The vehicle passage record data in Kafka is as follows:

[0049]

[0050]

[0051] Extract the license plate number from the vehicle passage record data, i.e., extract the pass_plate_no from Kafka. Determine if the extracted license plate number matches the set regular expression. If it matches, retain the vehicle's passage record data; otherwise, discard it. Since vehicle management offices in different regions usually only store vehicle registration information for their own region, it is necessary to set corresponding regular expressions for different regions and use regular expressions to filter the vehicle passage record data to ensure that the vehicle's passage record data can be compared with the vehicle data stored by the vehicle management office later.

[0052] Retrieve vehicle registration information stored in the vehicle management office. The vehicle registration information includes license plate number cphm, license plate type cpzl, and other information. Put the license plate number and license plate type (i.e., cphm-cpzl) into Redis as the key, and the other information as the value, such as the vehicle body color.

[0053] In some embodiments of the present invention, the step of comparing the vehicle's vehicle passage record data with the vehicle data stored in the vehicle management office to obtain suspected fake license plate data that is initially identified as fake, and storing the suspected fake license plate data in a first index specifically includes:

[0054] Based on the obtained vehicle passage records, a query is performed in Redis using the license plate number (pass_plate_no) and license plate type (pass_plate_type) as keys to check if corresponding data exists. If corresponding data exists, the license plate is considered genuine, and the vehicle passage record data is not stored in the first index `its_pre_fake`. If no corresponding data exists, the license plate is considered fake, and the vehicle passage record data is stored in the first index `its_pre_fake` in Elasticsearch. This generates a preliminary list of suspected fake license plates, `fakeList`. The fields of the first index include primary key, license plate number, license plate type, passage time, location ID, device ID, and region ID, as detailed below:

[0055]

[0056]

[0057] In some embodiments of the present invention, the step of querying all vehicle passage records of a vehicle within a preset time period based on vehicle passage record data to obtain the vehicle's trajectory data specifically includes:

[0058] The system retrieves a list of suspected fake license plate data (fakeList) from the first index (its_pre_fake). It then iterates through the suspected fake license plate data in fakeList, using the license plate number (cphm) and license plate type (cpzl) to query the vehicle passage records within a first preset time period before the passage time (gcsj) as the upbound vehicle passage trajectory list1, and the vehicle passage records within a second preset time period after the passage time as the downbound vehicle passage trajectory list2, thus obtaining the vehicle's trajectory data. The first preset time period and the second preset time period are equal. In specific implementations, the first and second preset time periods can be set according to actual needs. For example, using the passage time of the retrieved suspected fake license plate data as the intermediate time, the system queries the vehicle passage records within the first 3 hours as the upbound vehicle passage trajectory list1 and the vehicle passage records within the last 3 hours as the downbound vehicle passage trajectory list2.

[0059] Meanwhile, if the number of points for both the upward and downward vehicle passing trajectories is zero, the suspected fake license plate data will be deleted from the suspected fake license plate data list in the first index. Since the current coverage of checkpoints means that the probability of a vehicle being captured only once while driving on the road is very small, if the number of points for both the upward and downward vehicle passing trajectories is zero, it means that there is a high probability that the license plate is a data point of incorrect recognition, and the suspected fake license plate data can be directly deleted from the suspected fake license plate data list in the first index.

[0060] In some embodiments of the present invention, the step of calculating the joint recognition rate of checkpoints based on vehicle trajectory data and filtering suspected fake license plate data based on the joint recognition rate of checkpoints specifically includes:

[0061] When the total number of points on the upbound and downbound vehicle trajectories is greater than or equal to a preset value, the joint checkpoint recognition rate Score1 = N / M is calculated, where M is a constant set according to the checkpoint coverage rate, and N is the total number of points on the upbound and downbound vehicle trajectories. For example, if M is 6, and for a suspected fake license plate data, the total number of points on the upbound and downbound vehicle trajectories is 5, then the joint checkpoint recognition rate Score1 = 5 / 6. Preferably, the preset value is 4, and M is 6. That is, when the total number of points on the upbound and downbound vehicle trajectories is greater than or equal to 4, it indicates that the checkpoint coverage rate in the area is relatively high. Through actual experiments, a value of M of 6 can obtain relatively ideal results. Of course, the present invention is not limited to this. In specific implementation, the preset value and the value of M can be adjusted according to the actual checkpoint coverage rate, etc.

[0062] The system determines whether the calculated joint recognition rate score1 is greater than or equal to a preset first threshold. If so, the suspected fake license plate data is classified as high-probability fake license plate data and stored in the second index its_fake. The fields of the second index include primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition, as detailed below:

[0063]

[0064] If not, the suspected fake license plate data is deleted from the suspected fake license plate data list of the first index its_pre_fake. Preferably, the first threshold is 0.8, that is, if the calculated joint recognition rate score1 of the checkpoint is greater than or equal to 0.8, it indicates that the suspected fake license plate data has a relatively high probability of being a fake license plate, and at this time, the suspected fake license plate data is stored in the second index its_fake; of course, the present invention is not limited to this, and the value of the first threshold can be adjusted according to actual needs in specific implementation.

[0065] In some embodiments of the present invention, the step of calculating the probability of trajectory reasonableness based on vehicle trajectory data and location association configuration table, and filtering suspected fake license plate data based on the probability of trajectory reasonableness specifically includes:

[0066] When the total number of points in the upbound and downbound vehicle trajectories is less than a preset value, based on the point ID and upbound vehicle trajectory in the suspected fake license plate data, an upbound point association table corresponding to the upbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained; simultaneously, based on the point ID and downbound vehicle trajectory in the suspected fake license plate data, a downbound point association table corresponding to the downbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained, wherein the association probability of each point in the upbound and downbound point association tables is queried from the point association configuration table;

[0067] The trajectory rationality probability Score2 is calculated as A / C1 + B / C2, where A represents the sum of the association probabilities of each point in the uplink point association table, C1 represents the number of points in the uplink point association table, B represents the association probability of each point in the downlink point association table, and C2 represents the number of points in the downlink point association table.

[0068] The algorithm determines whether the calculated probability of the trajectory (score2) is greater than or equal to a preset second threshold. If so, the suspected fake license plate data is classified as high-probability fake and stored in the second index (its_fake). The fields of the second index include the primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition, as detailed below:

[0069]

[0070] If not, the suspected fake license plate data is deleted from the suspected fake license plate data list of the first index `its_pre_fake`. Preferably, the second threshold is 0.8, that is, if the calculated trajectory reasonableness probability `score2` is greater than or equal to 0.8, it indicates that the suspected fake license plate data has a relatively high probability of being fake, and at this time, the suspected fake license plate data is stored in the second index `its_fake`. Of course, the present invention is not limited to this, and the value of the first threshold can be adjusted according to actual needs in specific implementation. By comparing the first index `its_pre_fake` and the second index `its_fake`, it can be found that the second index `its_fake` has an additional `score` field compared to the first index `its_pre_fake`. This allows the front-end to sort according to the score, and fake license plate data can be manually identified.

[0071] The calculation of the probability of trajectory rationality will be further explained through a specific example below:

[0072] Based on the location ID (recorded as fake_dwid) and the upstream vehicle trajectory list1 of the suspected fake license plate data, the upstream location association table is as follows:

[0073]

[0074] Based on the location ID (recorded as fake_dwid) and downlink vehicle trajectory list2 of the suspected fake license plate data, the following downlink location association table was obtained:

[0075]

[0076] Then, the probability of trajectory rationality is score2 = (sp1 + sp2 + ... + spi) / i + (xp1 + xp2 + ... + xpj) / j, where i represents the number of points in the uplink point association table and j represents the number of points in the downlink point association table.

[0077] In practical implementation, for areas with high checkpoint coverage, the joint checkpoint recognition rate (Score1) can be directly calculated for filtering; for areas with low checkpoint coverage, the trajectory rationality probability can be calculated for filtering; of course, if necessary, the trajectory rationality probability can also be calculated for filtering in areas with high checkpoint coverage.

[0078] In summary, this invention utilizes the joint recognition rate of multiple checkpoints and the probability of trajectory reasonableness as a method for filtering license plate recognition errors. Specifically, it involves: acquiring preliminary identified fake license plate data: comparing vehicle passage record data with vehicle data stored in the vehicle management office to obtain a set of data initially identified as fake license plates. This data may contain license plate recognition errors due to various reasons, requiring further verification and filtering; acquiring trajectory data: for fake license plate data initially identified as fake, querying the vehicle's passage record data for the hours before and after the vehicle's passage to obtain the vehicle's trajectory data; calculating the joint recognition rate of checkpoints: for the vehicle's trajectory data, on the one hand, the accuracy of license plate recognition is judged by the joint recognition rate of multiple checkpoints. If the license plate passes through multiple checkpoints in the hours before and after the passage, and the trajectory is reasonable enough to identify the same vehicle, then the accuracy of license plate recognition is determined by combining the recognition rates of these checkpoints. The recognition rate can improve the confidence that the license plate is correctly identified; trajectory reasonableness probability calculation: On the other hand, the reasonableness probability of the vehicle trajectory can be calculated through statistical analysis. A reasonable vehicle trajectory should conform to traffic patterns and road networks. For example, a vehicle is unlikely to appear in two far apart locations in a short period of time. Combined with the pre-configured point association table, the reasonableness probability of the vehicle trajectory can be calculated to help determine the possibility of license plate recognition errors; comprehensive judgment of license plate error probability: By using the combined recognition rate of the checkpoint and the trajectory reasonableness probability, the error probability of each fake license plate data initially identified as fake can be obtained. According to the set threshold, the error recognition results with higher probabilities are filtered out, thereby obtaining a more accurate license plate recognition result. This invention utilizes data from multiple checkpoints and vehicle trajectory information, combining checkpoint joint recognition rate and trajectory reasonable probability to filter out fake license plate data initially identified as fake. This improves the reliability and accuracy of license plate recognition and reduces problems caused by misidentification. Through trajectory data analysis, abnormal trajectories across checkpoints can be detected, further improving the screening effect for fake license plates. Simultaneously, it can handle license plate recognition errors caused by specific scenarios, vehicle movement states, or shooting conditions, enhancing the method's adaptability and robustness.

[0079] Based on the same inventive concept, this application also provides an apparatus corresponding to the method in Embodiment 1, as detailed in Embodiment 2.

[0080] Example 2

[0081] This embodiment provides a license plate recognition error filtering device, such as... Figure 2 As shown, it includes a preliminary comparison module and a filtering module;

[0082] The preliminary comparison module is used to acquire vehicle passage record data and vehicle data stored in the vehicle management office, compare the vehicle passage record data with the vehicle data stored in the vehicle management office, obtain suspected fake license plate data that is initially identified as fake, and store the suspected fake license plate data in the first index; since the vehicle passage record data is collected and recorded by the license plate recognition system set up at the checkpoint, and the license plate recognition system may face various problems in actual use (such as image blurring, insufficient lighting, reflection, occlusion, etc.), it is difficult to avoid license plate recognition errors. Therefore, it is necessary to further filter the suspected fake license plate data that is initially identified as fake.

[0083] The filtering module is used to query all vehicle passage records within a preset time period based on vehicle passage record data for suspected fake license plate data that is initially identified as fake license plates, thereby obtaining the vehicle's trajectory data; calculate the joint recognition rate of checkpoints based on the vehicle trajectory data, and filter the suspected fake license plate data based on the joint recognition rate of checkpoints, thereby improving the reliability and accuracy of license plate recognition.

[0084] In this invention, the filtering module is further used to calculate the probability of trajectory rationality based on vehicle trajectory data and a point-of-origin association configuration table, and to filter suspected fake license plate data based on the trajectory rationality probability, thereby further improving the reliability and accuracy of license plate recognition. During initialization, the point-of-origin association configuration table calculates the probability distribution of the uplink and downlink points corresponding to each point by analyzing vehicle trajectory data over a period of time. The point-of-origin association configuration table is as follows:

[0085] Point ID Related point ID Related point types Association probability 11 22 Up 0.5 11 33 Up 0.3 11 44 Downward 0.7 11 none Up 0.2 11 none Downward 0.2 …

[0086] In a specific implementation of this invention, the point-to-point configuration table can be initialized by calculating the probability distribution of the uplink and downlink points corresponding to each point by calculating the vehicle trajectory data within a month. After initialization, it can also be manually adjusted. Specifically, when initializing the location association configuration table, vehicle passage trajectory data with license plates within the past month can be stored in the database. The SQL window functions `lag()` and `lead()` are then used to calculate the number of associations between each checkpoint and its downstream checkpoints, and the number of associations between each checkpoint and its upstream checkpoints within the past month, generating two intermediate result tables. These two intermediate result tables are then used to calculate the location association configuration table. For example, for checkpoint 1, which has checkpoints 2 and 3 downstream and checkpoint 4 upstream, the specific calculation involves first adding the association counts between checkpoint 2 and checkpoint 1, checkpoint 3 and checkpoint 1, and checkpoint 4 and checkpoint 1 to obtain the total association count for checkpoint 1. Then, the association count between checkpoint 2 and checkpoint 1 is divided by the total association count for checkpoint 1 to obtain the association probability between checkpoint 2 and checkpoint 1.

[0087] In some embodiments of the present invention, the acquisition of vehicle passage record data and vehicle data stored in the vehicle management office specifically includes:

[0088] Vehicle passage record data is retrieved from Kafka. The vehicle passage record data in Kafka is as follows:

[0089]

[0090]

[0091] Extract the license plate number from the vehicle passage record data, i.e., extract the pass_plate_no from Kafka. Determine if the extracted license plate number matches the set regular expression. If it matches, retain the vehicle's passage record data; otherwise, discard it. Since vehicle management offices in different regions usually only store vehicle registration information for their own region, it is necessary to set corresponding regular expressions for different regions and use regular expressions to filter the vehicle passage record data to ensure that the vehicle's passage record data can be compared with the vehicle data stored by the vehicle management office later.

[0092] Retrieve vehicle registration information stored in the vehicle management office. The vehicle registration information includes license plate number cphm, license plate type cpzl, and other information. Put the license plate number and license plate type (i.e., cphm-cpzl) into Redis as the key, and the other information as the value, such as the vehicle body color.

[0093] In some embodiments of the present invention, the step of comparing the vehicle's vehicle passage record data with the vehicle data stored in the vehicle management office to obtain suspected fake license plate data that is initially identified as fake, and storing the suspected fake license plate data in a first index specifically includes:

[0094] Based on the obtained vehicle passage records, a query is performed in Redis using the license plate number (pass_plate_no) and license plate type (pass_plate_type) as keys to check if corresponding data exists. If corresponding data exists, the license plate is considered genuine, and the vehicle passage record data is not stored in the first index `its_pre_fake`. If no corresponding data exists, the license plate is considered fake, and the vehicle passage record data is stored in the first index `its_pre_fake` in Elasticsearch. This generates a preliminary list of suspected fake license plates, `fakeList`. The fields of the first index include primary key, license plate number, license plate type, passage time, location ID, device ID, and region ID, as detailed below:

[0095]

[0096]

[0097] In some embodiments of the present invention, the step of querying all vehicle passage records of a vehicle within a preset time period based on vehicle passage record data to obtain the vehicle's trajectory data specifically includes:

[0098] The system retrieves a list of suspected fake license plate data (fakeList) from the first index (its_pre_fake). It then iterates through the suspected fake license plate data in fakeList, using the license plate number (cphm) and license plate type (cpzl) to query the vehicle passage records within a first preset time period before the passage time (gcsj) as the upbound vehicle passage trajectory list1, and the vehicle passage records within a second preset time period after the passage time as the downbound vehicle passage trajectory list2, thus obtaining the vehicle's trajectory data. The first preset time period and the second preset time period are equal. In specific implementations, the first and second preset time periods can be set according to actual needs. For example, using the passage time of the retrieved suspected fake license plate data as the intermediate time, the system queries the vehicle passage records within the first 3 hours as the upbound vehicle passage trajectory list1 and the vehicle passage records within the last 3 hours as the downbound vehicle passage trajectory list2.

[0099] Meanwhile, if the number of points for both the upward and downward vehicle passing trajectories is zero, the suspected fake license plate data will be deleted from the suspected fake license plate data list in the first index. Since the current coverage of checkpoints means that the probability of a vehicle being captured only once while driving on the road is very small, if the number of points for both the upward and downward vehicle passing trajectories is zero, it means that there is a high probability that the license plate is a data point of incorrect recognition, and the suspected fake license plate data can be directly deleted from the suspected fake license plate data list in the first index.

[0100] In some embodiments of the present invention, the step of calculating the joint recognition rate of checkpoints based on vehicle trajectory data and filtering suspected fake license plate data based on the joint recognition rate of checkpoints specifically includes:

[0101] When the total number of points on the upbound and downbound vehicle trajectories is greater than or equal to a preset value, the joint checkpoint recognition rate Score1 = N / M is calculated, where M is a constant set according to the checkpoint coverage rate, and N is the total number of points on the upbound and downbound vehicle trajectories. For example, if M is 6, and for a suspected fake license plate data, the total number of points on the upbound and downbound vehicle trajectories is 5, then the joint checkpoint recognition rate Score1 = 5 / 6. Preferably, the preset value is 4, and M is 6. That is, when the total number of points on the upbound and downbound vehicle trajectories is greater than or equal to 4, it indicates that the checkpoint coverage rate in the area is relatively high. Through actual experiments, a value of M of 6 can obtain relatively ideal results. Of course, the present invention is not limited to this. In specific implementation, the preset value and the value of M can be adjusted according to the actual checkpoint coverage rate, etc.

[0102] The system determines whether the calculated joint recognition rate score1 is greater than or equal to a preset first threshold. If so, the suspected fake license plate data is classified as high-probability fake license plate data and stored in the second index its_fake. The fields of the second index include primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition, as detailed below:

[0103]

[0104] If not, the suspected fake license plate data is deleted from the suspected fake license plate data list of the first index its_pre_fake. Preferably, the first threshold is 0.8, that is, if the calculated joint recognition rate score1 of the checkpoint is greater than or equal to 0.8, it indicates that the suspected fake license plate data has a relatively high probability of being a fake license plate, and at this time, the suspected fake license plate data is stored in the second index its_fake; of course, the present invention is not limited to this, and the value of the first threshold can be adjusted according to actual needs in specific implementation.

[0105] In some embodiments of the present invention, the step of calculating the probability of trajectory reasonableness based on vehicle trajectory data and location association configuration table, and filtering suspected fake license plate data based on the probability of trajectory reasonableness specifically includes:

[0106] When the total number of points in the upbound and downbound vehicle trajectories is less than a preset value, based on the point ID and upbound vehicle trajectory in the suspected fake license plate data, an upbound point association table corresponding to the upbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained; simultaneously, based on the point ID and downbound vehicle trajectory in the suspected fake license plate data, a downbound point association table corresponding to the downbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained, wherein the association probability of each point in the upbound and downbound point association tables is queried from the point association configuration table;

[0107] The trajectory rationality probability Score2 is calculated as A / C1 + B / C2, where A represents the sum of the association probabilities of each point in the uplink point association table, C1 represents the number of points in the uplink point association table, B represents the association probability of each point in the downlink point association table, and C2 represents the number of points in the downlink point association table.

[0108] The algorithm determines whether the calculated probability of the trajectory (score2) is greater than or equal to a preset second threshold. If so, the suspected fake license plate data is classified as high-probability fake and stored in the second index (its_fake). The fields of the second index include the primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition, as detailed below:

[0109]

[0110] If not, the suspected fake license plate data is deleted from the suspected fake license plate data list of the first index `its_pre_fake`. Preferably, the second threshold is 0.8, that is, if the calculated trajectory reasonableness probability `score2` is greater than or equal to 0.8, it indicates that the suspected fake license plate data has a relatively high probability of being fake, and at this time, the suspected fake license plate data is stored in the second index `its_fake`. Of course, the present invention is not limited to this, and the value of the first threshold can be adjusted according to actual needs in specific implementation. By comparing the first index `its_pre_fake` and the second index `its_fake`, it can be found that the second index `its_fake` has an additional `score` field compared to the first index `its_pre_fake`. This allows the front-end to sort according to the score, and fake license plate data can be manually identified.

[0111] The calculation of the probability of trajectory rationality will be further explained through a specific example below:

[0112] Based on the location ID (recorded as fake_dwid) and the upstream vehicle trajectory list1 of the suspected fake license plate data, the upstream location association table is as follows:

[0113]

[0114] Based on the location ID (recorded as fake_dwid) and downlink vehicle trajectory list2 of the suspected fake license plate data, the following downlink location association table was obtained:

[0115]

[0116] Then, the probability of trajectory rationality is score2 = (sp1 + sp2 + ... + spi) / i + (xp1 + xp2 + ... + xpj) / j, where i represents the number of points in the uplink point association table and j represents the number of points in the downlink point association table.

[0117] In practical implementation, for areas with high checkpoint coverage, the joint checkpoint recognition rate (Score1) can be directly calculated for filtering; for areas with low checkpoint coverage, the trajectory rationality probability can be calculated for filtering; of course, if necessary, the trajectory rationality probability can also be calculated for filtering in areas with high checkpoint coverage.

[0118] Based on the same inventive concept, this application provides an electronic device embodiment corresponding to Embodiment 1, as detailed in Embodiment 3.

[0119] Example 3

[0120] This embodiment provides an electronic device, such as... Figure 3 As shown, it includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it can implement any of the embodiments in Example 1.

[0121] Since the electronic device described in this embodiment is the device used to implement the method in Embodiment 1 of this application, those skilled in the art can understand the specific implementation method and various variations of the electronic device in this embodiment based on the method described in Embodiment 1 of this application. Therefore, how the electronic device implements the method in the embodiment of this application will not be described in detail here. Any device used by those skilled in the art to implement the method in the embodiment of this application falls within the scope of protection of this application.

[0122] Based on the same inventive concept, this application provides a storage medium corresponding to Embodiment 1, as detailed in Embodiment 4.

[0123] Example 4

[0124] This embodiment provides a computer-readable storage medium, such as... Figure 4 As shown, a computer program is stored thereon, which, when executed by a processor, can implement any of the embodiments in Example 1.

[0125] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0126] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus, and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0127] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0128] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0129] While specific embodiments of the present invention have been described above, those skilled in the art should understand that the specific embodiments described are merely illustrative and not intended to limit the scope of the present invention. Equivalent modifications and variations made by those skilled in the art in accordance with the spirit of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for filtering license plate recognition errors, characterized in that: The method includes the following steps: The system obtains vehicle passage record data and vehicle data stored in the vehicle management office. It compares the vehicle passage record data with the vehicle data stored in the vehicle management office to obtain suspected fake license plate data that is initially identified as fake license plates, and stores the suspected fake license plate data in the first index. For suspected fake license plate data initially identified as fake, the vehicle's trajectory data is obtained by querying all vehicle passage records within a preset time period based on the vehicle passage record data. Specifically, this includes: obtaining a list of suspected fake license plate data initially identified as fake in the first index; iteratively retrieving suspected fake license plate data from the suspected fake license plate data list, and querying the vehicle passage records within a first preset time period before the passage time as the upward passage trajectory and the vehicle passage records within a second preset time period after the passage time as the downward passage trajectory, based on the license plate number and license plate type in the suspected fake license plate data, thereby obtaining the vehicle's trajectory data. The first preset time period and the second preset time period are equal; calculating the joint recognition rate of the checkpoints based on the vehicle's trajectory data, specifically including calculating the joint recognition rate of the checkpoints Score1=N / M when the total number of points in the upward and downward passage trajectories is greater than or equal to a preset value, where M is a constant set according to the checkpoint coverage rate, and N is the total number of points in the upward and downward passage trajectories, and filtering the suspected fake license plate data based on the joint recognition rate of the checkpoints; It also includes calculating the probability of trajectory rationality based on vehicle trajectory data and point association configuration table. Specifically, when the total number of points in the upbound and downbound vehicle trajectories is less than a preset value, based on the point ID in the suspected fake license plate data and the upbound vehicle trajectory, the upbound point association table corresponding to the upbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained; at the same time, based on the point ID in the suspected fake license plate data and the downbound vehicle trajectory, the downbound point association table corresponding to the downbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained. The association probability of each point in the upbound and downbound point association tables is queried from the point association configuration table. The trajectory rationality probability Score2 = A / C1 + B / C2 is calculated, where A represents the sum of the association probabilities of each point in the uplink point association table, C1 represents the number of points in the uplink point association table, B represents the sum of the association probabilities of each point in the downlink point association table, and C2 represents the number of points in the downlink point association table; and suspected fake license plate data is filtered based on the trajectory rationality probability.

2. The method for filtering license plate recognition errors according to claim 1, characterized in that: During initialization, the point association configuration table calculates the probability distribution of the uplink and downlink points corresponding to each point by analyzing vehicle trajectory data over a period of time.

3. The method for filtering license plate recognition errors according to claim 1, characterized in that: The acquisition of vehicle passage record data and vehicle data stored in the vehicle management office specifically includes: Retrieve vehicle passage records from Kafka, extract the license plate numbers from the vehicle passage records, and determine whether the extracted license plate numbers match the set regular expression. If they match, keep the vehicle passage records; otherwise, discard the vehicle passage records. Retrieve vehicle registration information stored in the vehicle management office. The vehicle registration information includes license plate number, license plate type and other information. Put the license plate number and license plate type into Redis as the key and the other information as the value. The process of comparing vehicle passage record data with vehicle data stored in the vehicle management office to obtain suspected fake license plate data initially identified as fake, and storing the suspected fake license plate data in the first index, specifically includes: Based on the obtained vehicle passage records, the system queries Redis using the license plate number and license plate type as keys to check if corresponding data exists. If corresponding data exists, the license plate is considered genuine; otherwise, it is considered fake. The vehicle passage records are then stored in the first index of Elasticsearch, thus obtaining a preliminary list of suspected fake license plate data. The fields of the first index include primary key, license plate number, license plate type, passage time, location ID, device ID, and region ID.

4. The method for filtering license plate recognition errors according to claim 2, characterized in that: If the number of points for both the upward and downward vehicle passage trajectories is zero, then the suspected fake license plate data will be deleted from the suspected fake license plate data list of the first index.

5. The license plate recognition error filtering method according to claim 4, characterized in that: The filtering of suspected counterfeit license plate data based on the joint recognition rate of checkpoints specifically includes: If the calculated joint recognition rate score1 is greater than or equal to the preset first threshold, the suspected fake license plate data is classified as high-probability fake license plate data and stored in the second index. The fields of the second index include primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition. If not, the suspected fake license plate data is deleted from the suspected fake license plate data list in the first index.

6. The license plate recognition error filtering method according to claim 4, characterized in that: The filtering of suspected counterfeit license plate data based on the probability of trajectory reasonableness specifically includes: If the calculated probability score2 of the trajectory is greater than or equal to the preset second threshold, the suspected fake license plate data is classified as high-probability fake license plate data and stored in the second index. The fields of the second index include primary key, license plate number, license plate type, vehicle passage time, location ID, device ID, region ID, and the probability of correct recognition. If not, the suspected fake license plate data is deleted from the suspected fake license plate data list of the first index.

7. A license plate recognition error filtering device, characterized in that: Includes a preliminary comparison module and a filtering module; The preliminary comparison module is used to obtain vehicle passage record data and vehicle data stored in the vehicle management office, compare the vehicle passage record data with the vehicle data stored in the vehicle management office, obtain suspected fake license plate data that is initially identified as fake license plates, and store the suspected fake license plate data in the first index. The filtering module is used to query all vehicle passage records within a preset time period based on vehicle passage record data for suspected fake license plate data that is initially identified as fake, thereby obtaining the vehicle's trajectory data. Specifically, it includes: obtaining a list of suspected fake license plate data that is initially identified as fake in the first index; cyclically retrieving suspected fake license plate data from the suspected fake license plate data list, and querying vehicle passage records within a first preset time period before the passage time as the upward vehicle passage trajectory and vehicle passage records within a second preset time period after the passage time as the downward vehicle passage trajectory based on the license plate number and license plate type in the suspected fake license plate data, thereby obtaining the vehicle's trajectory data, where the first preset time period and the second preset time period are equal; calculating the joint recognition rate of the checkpoint based on the vehicle's trajectory data, specifically including calculating the joint recognition rate of the checkpoint Score1=N / M when the total number of points in the upward and downward vehicle passage trajectories is greater than or equal to a preset value, where M is a constant set according to the checkpoint coverage rate, N is the total number of points in the upward and downward vehicle passage trajectories, and filtering the suspected fake license plate data based on the joint recognition rate of the checkpoint. The filtering module is also used to calculate the trajectory rationality probability based on the vehicle's trajectory data and the point association configuration table, and to filter suspected fake license plate data based on the trajectory rationality probability. Specifically, when the total number of points in the upbound and downbound vehicle trajectories is less than a preset value, the upbound point association table corresponding to the upbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained based on the point ID and the upbound vehicle trajectory in the suspected fake license plate data; at the same time, the downbound point association table corresponding to the downbound vehicle trajectory starting from the point ID of the suspected fake license plate data is obtained based on the point ID and the downbound vehicle trajectory in the suspected fake license plate data. The association probability of each point in the upbound and downbound point association tables is queried from the point association configuration table. The trajectory rationality probability Score2 = A / C1 + B / C2 is calculated, where A represents the sum of the association probabilities of each point in the uplink point association table, C1 represents the number of points in the uplink point association table, B represents the sum of the association probabilities of each point in the downlink point association table, and C2 represents the number of points in the downlink point association table; and suspected fake license plate data is filtered based on the trajectory rationality probability.

8. The license plate recognition error filtering device according to claim 7, characterized in that: During initialization, the point association configuration table calculates the probability distribution of the uplink and downlink points corresponding to each point by analyzing vehicle trajectory data over a period of time.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 6.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 6.

Images