Method and device for auditing vehicle fee evasion

Abstract

The present application discloses a method and device for auditing vehicle toll evasion, which relates to the field of highway data verification and aims to solve the technical problem of low accuracy in auditing vehicle toll evasion, comprising: obtaining vehicle traffic data of multiple toll stations and multiple highway sections, and performing feature extraction on the vehicle traffic data to obtain a feature set of passing vehicles; based on preset traffic feature markers, segmenting the feature set of passing vehicles to obtain a feature set of passing vehicles and a feature set of road network vehicles; based on a preset feature difference scoring method, obtaining a first feature set of vehicle toll evasion according to the feature set of passing vehicles and the feature set of road network vehicles; based on a preset road network model, screening the first feature set of vehicle toll evasion to obtain a second feature set of vehicle toll evasion; and calculating the toll payable for each vehicle in the second feature set of vehicle toll evasion according to the vehicle traffic data, so as to audit vehicle toll evasion based on the toll payable.

Description

Technical Field

[0001] The present application relates to the field of highway data verification, and in particular to a method and device for verifying vehicle toll evasion. Background Art

[0002] With the continuous construction and development of expressways, the expressway network has closely connected the transportation of various provinces. This has led to the inspection and analysis of a large number of vehicles traveling across provinces, especially the inspection of vehicles evading fees among vehicles traveling across provinces. At present, there are two main difficulties in the inspection of vehicles evading fees across provinces: first, it is difficult to identify similar vehicles, which can easily lead to misjudgment and reduce the accuracy of the inspection; second, in the process of traveling across provinces, it is difficult to correctly identify vehicles that evade fees by using false license plates while excluding legal changes in vehicle license plates, making the inspection difficult. Therefore, how to improve the accuracy of the inspection of vehicles evading fees is still a problem that needs to be solved urgently in the existing technology. Summary of the Invention

[0003] The present application provides a method and device for auditing vehicle fee evasion to solve the technical problem of low accuracy in auditing vehicle fee evasion.

[0004] According to a first aspect of the embodiments of the present application, a method for auditing vehicle fee evasion and non-payment is provided, comprising:

[0005] Acquiring vehicle traffic data from multiple toll booths and multiple highway sections, and performing feature extraction on the vehicle traffic data to obtain a feature set of passing vehicles; wherein the multiple highway sections are located between the multiple toll booths; the vehicle traffic data includes vehicle traffic time, vehicle identification features, and multiple frames of vehicle traffic images;

[0006] Based on the preset passing feature markers, the passing vehicle feature set is segmented to obtain a passing vehicle feature set and a road network vehicle feature set;

[0007] Based on a preset feature difference scoring method, a first toll evasion vehicle feature set is obtained according to the transit vehicle feature set and the road network vehicle feature set;

[0008] Based on a preset road network model, the first feature set of vehicles that evade tolls is screened to obtain a second feature set of vehicles that evade tolls;

[0009] The highway toll payable for each vehicle in the second toll evasion vehicle feature set is calculated based on the vehicle traffic data, so as to conduct a vehicle toll evasion audit based on the toll payable toll.

[0010] This application first extracts features from vehicle traffic data to obtain a passing vehicle feature set, then divides the passing vehicle feature set to obtain a passing vehicle feature set and a road network vehicle feature set, and then obtains a first toll-evading vehicle feature set based on a feature difference scoring method. By comparing the feature differences between vehicles passing through toll stations and vehicles passing through expressway sections, the recognition accuracy of similar vehicles can be improved and the probability of misjudgment can be reduced. The first toll-evading vehicle feature set is then filtered based on a road network model to obtain a second toll-evading vehicle feature set. Based on the road network model, some normal passing vehicles that are not toll-evading can be filtered out, and then when calculating the highway tolls to be paid for each vehicle in the second toll-evading vehicle feature set for vehicle toll-evading audit, the accuracy of vehicle toll-evading audit can be improved.

[0011] In certain embodiments of the present application, extracting features from the vehicle traffic data to obtain a passing vehicle feature set specifically includes:

[0012] Segmenting a plurality of vehicle traffic images of each vehicle in the vehicle traffic data according to a deep learning-based image segmentation method to obtain vehicle color features, vehicle model features, and vehicle damage features of each vehicle;

[0013] Normalize and concatenate the vehicle color feature, vehicle model feature, and vehicle damage feature of each vehicle to obtain the first comprehensive feature of each vehicle;

[0014] Based on a preset standard, the first comprehensive feature of each vehicle is normalized to obtain a second comprehensive feature of each vehicle;

[0015] A passing vehicle feature set is obtained based on the second comprehensive feature of each vehicle, the vehicle identification feature and the vehicle passing time.

[0016] This application first segments the vehicle traffic image according to the deep learning-based image segmentation method to obtain vehicle color features, vehicle model features and vehicle damage features, and then performs normalization, splicing and standardization operations in succession to obtain a feature set of passing vehicles, which can facilitate the subsequent calculation of feature differences and improve the efficiency and accuracy of feature difference calculation.

[0017] In certain embodiments of the present application, the preset feature difference scoring method, based on the transit vehicle feature set and the road network vehicle feature set, obtains a first toll evasion vehicle feature set, specifically including:

[0018] Sequentially selecting each vehicle in the transit vehicle feature set as the current first vehicle, and obtaining an audit result of the current first vehicle according to a preset vehicle fee evasion audit method, until all vehicles in the transit vehicle feature set are selected, and obtaining a first fee evasion vehicle feature set according to the audit result;

[0019] The method for auditing vehicle fee evasion is as follows:

[0020] Calculating a feature difference score between the current first vehicle and each vehicle in the road network vehicle feature set based on a preset feature difference scoring method;

[0021] If any one of the multiple feature difference scores of the current first vehicle reaches a preset threshold, the license plate numbers of the current first vehicle and the current second vehicle are obtained based on several frames of vehicle traffic images of the current first vehicle and several frames of vehicle traffic images of the current second vehicle; wherein the current second vehicle is a vehicle in the road network vehicle feature set corresponding to the feature difference score that reaches the preset threshold;

[0022] If the license plate number of the current first vehicle is different from the license plate number of the current second vehicle, the vehicle registration information of the current first vehicle is obtained based on the vehicle identification characteristics of the current first vehicle, and the audit result of the current first vehicle is obtained based on the vehicle registration information.

[0023] This application first selects the current first vehicle in the passing vehicle feature set, then obtains the feature difference score of the current first vehicle and each vehicle in the road network vehicle feature set, and obtains the corresponding license plate numbers of the current first vehicle and the current second vehicle when any one of the multiple feature difference scores reaches a preset threshold and compares them. It can first exclude two vehicles with similar features but are actually different, determine whether the current first vehicle and the current second vehicle are the same vehicle, and then obtain the vehicle registration information based on the comparison result of the license plate numbers, and then obtain the audit result. It can exclude vehicles with legally changed license plates, improve the recognition accuracy of vehicles that evade fees, and thus improve the accuracy of auditing vehicles that evade fees.

[0024] In certain embodiments of the present application, the feature difference scoring method is specifically:

[0025] Sequentially selecting each vehicle in the road network vehicle feature set as a current scoring vehicle, and calculating a feature difference score between the current first vehicle and the current scoring vehicle according to a preset first scoring method, until all vehicles in the road network vehicle feature set have been selected;

[0026] The first scoring method is specifically:

[0027] Based on a similarity algorithm, multiple similarities between the current first vehicle and the currently rated vehicle are calculated; wherein the multiple similarities are based on vehicle color features, vehicle model features, vehicle damage features, and vehicle identification features of the current first vehicle and the currently rated vehicle;

[0028] A feature difference score between the current first vehicle and the current scoring vehicle is obtained based on the multiple similarities.

[0029] This application first selects the currently scored vehicle from the road network vehicle feature set, and then based on the similarity algorithm, calculates multiple similarities between the current first vehicle and the currently scored vehicle, including vehicle color features, vehicle model features, vehicle damage features and vehicle identification features, and determines the feature difference score based on the multiple similarities. This can improve the accuracy of the feature difference score, and then more accurately determine whether the current first vehicle and the currently scored vehicle are the same vehicle, thereby improving the accuracy of identifying similar vehicles.

[0030] In certain embodiments of the present application, the first toll-evading vehicle feature set is screened based on a preset road network model to obtain a second toll-evading vehicle feature set, specifically including:

[0031] Sequentially selecting each vehicle in the first vehicle feature set as a first identification vehicle, and obtaining an identification result of the first identification vehicle according to a preset vehicle identification method for evading tolls, until all vehicles in the first vehicle feature set have been selected;

[0032] The method for determining whether a vehicle has evaded or missed charges is as follows:

[0033] Acquire multiple toll station passage data corresponding to the first identified vehicle; wherein the toll station passage data includes toll station location data, vehicle passing time, and vehicle toll fees;

[0034] Based on a preset road network model, according to the toll station location data and the vehicle's transit time, a first driving path of the first identified vehicle is obtained;

[0035] Based on the road network model and the preset path algorithm, and according to the toll station location data, a first theoretical path and a plurality of first theoretical fees for the first identified vehicle are obtained;

[0036] Obtaining a first vehicle identification result based on the first driving path and the first theoretical path;

[0037] Obtaining a second discrimination result of the first discrimination vehicle based on the vehicle toll and the plurality of first theoretical fees;

[0038] Obtaining a third discrimination result of the first discrimination vehicle according to the vehicle passing time;

[0039] According to the first discrimination result, the second discrimination result and the third discrimination result, a discrimination result of the first discrimination vehicle for toll evasion is obtained.

[0040] This application first selects the first judgment vehicle, then obtains the traffic data of multiple toll stations corresponding to the first judgment vehicle, and based on the road network model and path algorithm, obtains the first driving path, the first theoretical path and multiple first theoretical fees, and then obtains the first judgment result, the second judgment result and the third judgment result. This multi-angle judgment can improve the accuracy of the final toll evasion judgment, and thus improve the accuracy of the audit of vehicle toll evasion.

[0041] In certain embodiments of the present application, the calculating, based on the vehicle traffic data, the highway toll payable for each vehicle in the second toll evasion vehicle feature set specifically includes:

[0042] Sequentially selecting each vehicle in the second toll evasion vehicle feature set as the current toll-billing vehicle, and obtaining the toll payable for the current toll-billing vehicle according to a preset toll evasion billing method, until all vehicles in the second toll evasion vehicle feature set have been selected;

[0043] The method for charging evasion fees is as follows:

[0044] Determine the entry and exit data of the toll station corresponding to the current billing vehicle based on multiple vehicle travel times of the current billing vehicle;

[0045] If both the entry data and the exit data are not empty, the highway toll payable for the current billing vehicle is obtained based on the entry data and the exit data;

[0046] If either the entry data or the exit data is empty, the second theoretical path of the current billing vehicle is obtained based on the vehicle passage data of the current billing vehicle, and the highway toll that should be paid by the current billing vehicle is obtained based on the second theoretical path.

[0047] This application first selects the current billing vehicle, and then determines the corresponding toll station's entry and exit data based on the multiple vehicle travel times of the current billing vehicle, and calculates the highway toll that the current billing vehicle should pay based on the null value of the entry and exit data. This can improve the accuracy of calculating the highway toll that should be paid for vehicles that evade fees, and thereby improve the accuracy of auditing vehicles that evade fees.

[0048] According to a second aspect of the embodiments of the present application, there is provided a vehicle fee evasion audit device, comprising a vehicle feature extraction module, a feature set segmentation module, a first fee evasion set determination module, a second fee evasion set determination module, and a vehicle fee supplement calculation module;

[0049] The vehicle feature extraction module is configured to obtain vehicle traffic data from multiple toll booths and multiple highway sections, and perform feature extraction on the vehicle traffic data to obtain a traffic feature set; wherein the multiple highway sections are located between the multiple toll booths; the vehicle traffic data includes vehicle traffic time, vehicle identification features, and multiple frames of vehicle traffic images;

[0050] The feature set segmentation module is used to segment the passing vehicle feature set based on the preset passing feature markers to obtain the passing vehicle feature set and the road network vehicle feature set;

[0051] The first toll evasion set determination module is configured to obtain a first toll evasion vehicle feature set based on the transit vehicle feature set and the road network vehicle feature set based on a preset feature difference scoring method;

[0052] The second fee evasion set determination module is configured to filter the first fee evasion vehicle feature set based on a preset road network model to obtain a second fee evasion vehicle feature set;

[0053] The vehicle fee calculation module is used to calculate the highway fee that should be paid for each vehicle in the second toll evasion vehicle feature set based on the vehicle passage data, so as to conduct vehicle toll evasion audit based on the highway fee that should be paid.

[0054] In certain embodiments of the present application, the vehicle feature extraction module includes an image segmentation submodule, an image normalization submodule, a feature specification submodule, and a feature acquisition submodule;

[0055] The image segmentation submodule is used to segment several frames of vehicle traffic images of each vehicle in the vehicle traffic data according to an image segmentation method based on deep learning, so as to obtain vehicle color features, vehicle model features and vehicle damage features of each vehicle;

[0056] The image normalization submodule is used to normalize and splice the vehicle color features, vehicle model features, and vehicle damage features of each vehicle to obtain a first comprehensive feature of each vehicle;

[0057] The feature normalization submodule is configured to normalize the first comprehensive feature of each vehicle based on a preset standard to obtain a second comprehensive feature of each vehicle;

[0058] The feature acquisition submodule is used to obtain a passing vehicle feature set based on the second comprehensive feature of each vehicle, the vehicle identification feature and the vehicle passing time.

[0059] In certain embodiments of the present application, the first evasion fee set determination module includes a first evasion fee set determination submodule;

[0060] The first fee evasion set determination submodule is configured to sequentially select each vehicle in the transit vehicle feature set as the current first vehicle, obtain an audit result of the current first vehicle according to a preset vehicle fee evasion audit method, and obtain the first fee evasion vehicle feature set based on the audit result until all vehicles in the transit vehicle feature set have been selected.

[0061] The method for auditing vehicle fee evasion is as follows:

[0062] Calculating a feature difference score between the current first vehicle and each vehicle in the road network vehicle feature set based on a preset feature difference scoring method;

[0063] If any one of the multiple feature difference scores of the current first vehicle reaches a preset threshold, the license plate numbers of the current first vehicle and the current second vehicle are obtained based on several frames of vehicle traffic images of the current first vehicle and several frames of vehicle traffic images of the current second vehicle; wherein the current second vehicle is a vehicle in the road network vehicle feature set corresponding to the feature difference score that reaches the preset threshold;

[0064] If the license plate number of the current first vehicle is different from the license plate number of the current second vehicle, the vehicle registration information of the current first vehicle is obtained based on the vehicle identification characteristics of the current first vehicle, and the audit result of the current first vehicle is obtained based on the vehicle registration information.

[0065] In certain embodiments of the present application, the feature difference scoring method is specifically:

[0066] Sequentially selecting each vehicle in the road network vehicle feature set as a current scoring vehicle, and calculating a feature difference score between the current first vehicle and the current scoring vehicle according to a preset first scoring method, until all vehicles in the road network vehicle feature set have been selected;

[0067] The first scoring method is specifically:

[0068] Based on a similarity algorithm, multiple similarities between the current first vehicle and the currently rated vehicle are calculated; wherein the multiple similarities are based on vehicle color features, vehicle model features, vehicle damage features, and vehicle identification features of the current first vehicle and the currently rated vehicle;

[0069] A feature difference score between the current first vehicle and the current scoring vehicle is obtained based on the multiple similarities.

[0070] In certain embodiments of the present application, the second evasion fee set determination module includes a second evasion fee set determination submodule;

[0071] The second fee evasion set determination submodule is configured to sequentially select each vehicle in the first fee evasion vehicle feature set as a first discrimination vehicle, and obtain a fee evasion discrimination result for the first discrimination vehicle according to a preset vehicle fee evasion discrimination method, until all vehicles in the first fee evasion vehicle feature set have been selected;

[0072] The method for determining whether a vehicle has evaded or missed charges is as follows:

[0073] Acquire multiple toll station passage data corresponding to the first identified vehicle; wherein the toll station passage data includes toll station location data, vehicle passing time, and vehicle toll fees;

[0074] Based on a preset road network model, according to the toll station location data and the vehicle's transit time, a first driving path of the first identified vehicle is obtained;

[0075] Based on the road network model and the preset path algorithm, and according to the toll station location data, a first theoretical path and a plurality of first theoretical fees for the first identified vehicle are obtained;

[0076] Obtaining a first vehicle identification result based on the first driving path and the first theoretical path;

[0077] Obtaining a second discrimination result of the first discrimination vehicle based on the vehicle toll and the plurality of first theoretical fees;

[0078] Obtaining a third discrimination result of the first discrimination vehicle according to the vehicle passing time;

[0079] According to the first discrimination result, the second discrimination result and the third discrimination result, a discrimination result of the first discrimination vehicle for toll evasion is obtained.

[0080] In certain embodiments of the present application, the vehicle fee calculation module includes a vehicle fee calculation submodule;

[0081] The vehicle toll calculation submodule is configured to sequentially select each vehicle in the second toll evasion vehicle feature set as the current toll-charging vehicle, and obtain the toll payable for the current toll-charging vehicle according to a preset toll evasion billing method, until all vehicles in the second toll evasion vehicle feature set have been selected;

[0082] The method for charging evasion fees is as follows:

[0083] Determine the entry and exit data of the toll station corresponding to the current billing vehicle based on multiple vehicle travel times of the current billing vehicle;

[0084] If both the entry data and the exit data are not empty, the highway toll payable for the current billing vehicle is obtained based on the entry data and the exit data;

[0085] If either the entry data or the exit data is empty, the second theoretical path of the current billing vehicle is obtained based on the vehicle passage data of the current billing vehicle, and the highway toll that should be paid by the current billing vehicle is obtained based on the second theoretical path.

[0086] This application first extracts features from vehicle traffic data to obtain a passing vehicle feature set, then divides the passing vehicle feature set to obtain a passing vehicle feature set and a road network vehicle feature set, and then obtains a first toll-evading vehicle feature set based on a feature difference scoring method. By comparing the feature differences between vehicles passing through toll stations and vehicles passing through expressway sections, the recognition accuracy of similar vehicles can be improved and the probability of misjudgment can be reduced. The first toll-evading vehicle feature set is then filtered based on a road network model to obtain a second toll-evading vehicle feature set. Based on the road network model, some normal passing vehicles that are not toll-evading can be filtered out, and then when calculating the highway tolls to be paid for each vehicle in the second toll-evading vehicle feature set for vehicle toll-evading audit, the accuracy of vehicle toll-evading audit can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 : A flow chart of a method for auditing vehicle fee evasion shown in certain embodiments of the present application;

[0088] Figure 2 : A module structure diagram of a vehicle fee evasion and auditing device shown in certain embodiments of the present application. DETAILED DESCRIPTION

[0089] The embodiments of the present application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals represent the same or similar elements or elements with the same or similar functions from beginning to end. The embodiments described below in conjunction with the accompanying drawings are exemplary and are only used to explain some embodiments of the present application and should not be understood as limiting the embodiments of the present application. Based on the embodiments shown in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0090] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be understood to indicate or imply relative importance or implicitly indicate the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include at least one of such features. In the description of this application, unless otherwise clearly specified, "multiple" and "several" mean two or more.

[0091] Currently, there are two main difficulties in auditing vehicles that evade tolls: (1) it is difficult to identify similar vehicles, which can easily lead to misjudgment and thus reduce the accuracy of the audit; (2) during inter-provincial travel, it is difficult to exclude vehicles with legally changed license plates, which in turn increases the difficulty of correctly identifying vehicles that evade tolls by using false license plates, thus increasing the difficulty of auditing and reducing the accuracy of the audit. Therefore, how to improve the accuracy of vehicle toll evasion audits remains a problem that needs to be solved urgently in existing technologies.

[0092] Based on the above technical background, please refer to Figure 1 The embodiment of the present application provides a method for auditing vehicle fee evasion, including steps S101 to S105, each of which is specifically as follows:

[0093] Step S101: Acquire vehicle traffic data of multiple toll stations and multiple highway sections, and perform feature extraction on the vehicle traffic data to obtain a feature set of passing vehicles; wherein the multiple highway sections are located between the multiple toll stations; the vehicle traffic data includes vehicle traffic time, vehicle identification features and several frames of vehicle traffic images.

[0094] In certain embodiments of the present application, extracting features from the vehicle traffic data to obtain a passing vehicle feature set specifically includes:

[0095] Segmenting a plurality of vehicle traffic images of each vehicle in the vehicle traffic data according to a deep learning-based image segmentation method to obtain vehicle color features, vehicle model features, and vehicle damage features of each vehicle;

[0096] Normalize and concatenate the vehicle color feature, vehicle model feature, and vehicle damage feature of each vehicle to obtain the first comprehensive feature of each vehicle;

[0097] Based on a preset standard, the first comprehensive feature of each vehicle is normalized to obtain a second comprehensive feature of each vehicle;

[0098] A passing vehicle feature set is obtained based on the second comprehensive feature of each vehicle, the vehicle identification feature and the vehicle passing time.

[0099] In certain embodiments of the present application, the preferred segmentation model of the deep learning-based image segmentation method is a convolutional neural network, which includes but is not limited to R-CNN, Fast R-CNN and Faster R-CNN.

[0100] In certain embodiments of the present application, the preferred implementation of the vehicle color feature is a vector composed of the mean and variance of the RGB three channels of the vehicle body; the preferred implementation of the vehicle model feature is a vector composed of the aspect ratio of the vehicle body contour and the height ratio of the front of the vehicle to the body; the preferred implementation of the vehicle damage feature is a vector composed of the position coordinates of the vehicle body scratches and vehicle body dents obtained based on the recognition algorithm.

[0101] This application first segments the vehicle traffic image according to the deep learning-based image segmentation method to obtain vehicle color features, vehicle model features and vehicle damage features, and then performs normalization, splicing and standardization operations in succession to obtain a feature set of passing vehicles, which can facilitate the subsequent calculation of feature differences and improve the efficiency and accuracy of feature difference calculation.

[0102] Step S102: Based on the preset traffic feature markers, the traffic feature set is segmented to obtain a passing vehicle feature set and a road network vehicle feature set.

[0103] In certain embodiments of the present application, the traffic feature mark is a vehicle traffic section type, including a toll booth and a highway section.

[0104] Step S103: Based on a preset feature difference scoring method, a first toll evasion vehicle feature set is obtained according to the transit vehicle feature set and the road network vehicle feature set.

[0105] In certain embodiments of the present application, the preset feature difference scoring method, based on the transit vehicle feature set and the road network vehicle feature set, obtains a first toll evasion vehicle feature set, specifically including:

[0106] Sequentially selecting each vehicle in the transit vehicle feature set as the current first vehicle, and obtaining an audit result of the current first vehicle according to a preset vehicle fee evasion audit method, until all vehicles in the transit vehicle feature set are selected, and obtaining a first fee evasion vehicle feature set according to the audit result;

[0107] The method for auditing vehicle fee evasion is as follows:

[0108] Calculating a feature difference score between the current first vehicle and each vehicle in the road network vehicle feature set based on a preset feature difference scoring method;

[0109] If any one of the multiple feature difference scores of the current first vehicle reaches a preset threshold, the license plate numbers of the current first vehicle and the current second vehicle are obtained based on several frames of vehicle traffic images of the current first vehicle and several frames of vehicle traffic images of the current second vehicle; wherein the current second vehicle is a vehicle in the road network vehicle feature set corresponding to the feature difference score that reaches the preset threshold;

[0110] If the license plate number of the current first vehicle is different from the license plate number of the current second vehicle, the vehicle registration information of the current first vehicle is obtained based on the vehicle identification characteristics of the current first vehicle, and the audit result of the current first vehicle is obtained based on the vehicle registration information.

[0111] In certain embodiments of the present application, obtaining the audit result of the current first vehicle based on the vehicle registration information is specifically:

[0112] If there is no license plate change record in the vehicle registration information, the current audit result of the first vehicle is "toll evasion vehicle".

[0113] This application first selects the current first vehicle in the passing vehicle feature set, then obtains the feature difference score of the current first vehicle and each vehicle in the road network vehicle feature set, and obtains the corresponding license plate numbers of the current first vehicle and the current second vehicle when any one of the multiple feature difference scores reaches a preset threshold and compares them. It can first exclude two vehicles with similar features but are actually different, determine whether the current first vehicle and the current second vehicle are the same vehicle, and then obtain the vehicle registration information based on the comparison result of the license plate numbers, and then obtain the audit result. It can exclude vehicles with legally changed license plates, improve the recognition accuracy of vehicles that evade fees, and thus improve the accuracy of auditing vehicles that evade fees.

[0114] In certain embodiments of the present application, the feature difference scoring method is specifically:

[0115] Sequentially selecting each vehicle in the road network vehicle feature set as a current scoring vehicle, and calculating a feature difference score between the current first vehicle and the current scoring vehicle according to a preset first scoring method, until all vehicles in the road network vehicle feature set have been selected;

[0116] The first scoring method is specifically:

[0117] Based on a similarity algorithm, multiple similarities between the current first vehicle and the currently rated vehicle are calculated; wherein the multiple similarities are based on vehicle color features, vehicle model features, vehicle damage features, and vehicle identification features of the current first vehicle and the currently rated vehicle;

[0118] A feature difference score between the current first vehicle and the current scoring vehicle is obtained based on the multiple similarities.

[0119] In certain embodiments of the present application, preferred implementations of the similarity algorithm include but are not limited to a cosine similarity algorithm, a PSNR algorithm, a structural similarity (SSIM) algorithm, and a mean hash algorithm.

[0120] In certain embodiments of the present application, the feature difference score between the current first vehicle and the current scoring vehicle is obtained based on the multiple similarities, specifically:

[0121] A weighted sum of the multiple similarities under preset weights is used as a feature difference score between the current first vehicle and the current scoring vehicle.

[0122] In certain embodiments of the present application, the preferred implementation scheme of the preset weights is as follows: the weights of the similarities corresponding to vehicle color features, vehicle model features, vehicle damage features and vehicle identification features are 0.3, 0.4, 0.2 and 0.1 respectively.

[0123] This application first selects the currently scored vehicle from the road network vehicle feature set, and then based on the similarity algorithm, calculates multiple similarities between the current first vehicle and the currently scored vehicle, including vehicle color features, vehicle model features, vehicle damage features and vehicle identification features, and determines the feature difference score based on the multiple similarities. This can improve the accuracy of the feature difference score, and then more accurately determine whether the current first vehicle and the currently scored vehicle are the same vehicle, thereby improving the accuracy of identifying similar vehicles.

[0124] Step S104: Based on a preset road network model, the first feature set of toll-evading vehicles is screened to obtain a second feature set of toll-evading vehicles.

[0125] In certain embodiments of the present application, the first toll-evading vehicle feature set is screened based on a preset road network model to obtain a second toll-evading vehicle feature set, specifically including:

[0126] Sequentially selecting each vehicle in the first vehicle feature set as a first identification vehicle, and obtaining an identification result of the first identification vehicle according to a preset vehicle identification method for evading tolls, until all vehicles in the first vehicle feature set have been selected;

[0127] The method for determining whether a vehicle has evaded or missed charges is as follows:

[0128] Acquire multiple toll station passage data corresponding to the first identified vehicle; wherein the toll station passage data includes toll station location data, vehicle passing time, and vehicle toll fees;

[0129] Based on a preset road network model, according to the toll station location data and the vehicle's transit time, a first driving path of the first identified vehicle is obtained;

[0130] Based on the road network model and the preset path algorithm, and according to the toll station location data, a first theoretical path and a plurality of first theoretical fees for the first identified vehicle are obtained;

[0131] Obtaining a first vehicle identification result based on the first driving path and the first theoretical path;

[0132] Obtaining a second discrimination result of the first discrimination vehicle based on the vehicle toll and the plurality of first theoretical fees;

[0133] Obtaining a third discrimination result of the first discrimination vehicle according to the vehicle passing time;

[0134] According to the first discrimination result, the second discrimination result and the third discrimination result, a discrimination result of the first discrimination vehicle for toll evasion is obtained.

[0135] In certain embodiments of the present application, the first driving path of the first identified vehicle is obtained based on the preset road network model, the toll station location data and the vehicle transit time, and is specifically:

[0136] According to the toll station location data, a plurality of first toll stations corresponding to the first identified vehicle are obtained;

[0137] Based on a preset road network model, the multiple first toll stations are connected in order of the vehicle's passing time to obtain a first driving path of the first identified vehicle.

[0138] In certain embodiments of the present application, based on the road network model and the preset path algorithm, the first theoretical path and multiple first theoretical fees of the first identified vehicle are obtained according to the toll station location data, specifically:

[0139] According to the toll station location data, a plurality of second toll stations corresponding to the first identified vehicle are obtained;

[0140] Based on the road network model and the preset path algorithm, connecting the plurality of second toll stations to obtain a first theoretical path of the first identified vehicle;

[0141] A plurality of first theoretical fees are obtained based on the plurality of second toll stations and the first theoretical path.

[0142] In certain embodiments of the present application, the preset path algorithm includes but is not limited to an A* algorithm, a Dijkstra algorithm, and a simulated annealing algorithm.

[0143] In certain embodiments of the present application, the first vehicle identification result obtained based on the first driving path and the first theoretical path is specifically:

[0144] Calculating the path similarity between the first driving path and the first theoretical path according to a preset path similarity algorithm;

[0145] If the path similarity is lower than a preset path similarity threshold, the first identification result of the first identification vehicle is "toll evasion vehicle".

[0146] In certain embodiments of the present application, the path similarity algorithm includes but is not limited to a dynamic time normalization algorithm, a path similarity algorithm based on edit distance, and a path similarity algorithm based on the longest common subsequence.

[0147] In certain embodiments of the present application, the second identification result of the first identification vehicle is obtained based on the vehicle toll and the plurality of first theoretical fees, specifically:

[0148] If the ratio of the sum of the plurality of first theoretical fees to the sum of the plurality of vehicle tolls does not fall within the preset fee ratio range, the second determination result of the first determination of the vehicle is "abnormal vehicle".

[0149] In certain embodiments of the present application, the third discrimination result of the first discrimination vehicle is obtained based on the vehicle transit time, specifically:

[0150] According to the toll station location data, a plurality of third toll stations passed by the first identified vehicle are obtained;

[0151] Obtaining a plurality of preset travel times between the plurality of third toll booths;

[0152] If the ratio of the transit time of multiple vehicles to the preset passage time of the corresponding third toll station is greater than the preset passage time ratio threshold, the third judgment result of the first judgment vehicle is "toll evasion vehicle".

[0153] In certain embodiments of the present application, the first discrimination result of the vehicle evading tolls is obtained based on the first discrimination result, the second discrimination result, and the third discrimination result, specifically:

[0154] If any one of the first judgment result and the third judgment result does not meet the preset judgment conditions, and the second judgment result does not meet the preset judgment conditions, the first judgment result of the vehicle's fee evasion is "fee evasion".

[0155] This application first selects the first judgment vehicle, then obtains the traffic data of multiple toll stations corresponding to the first judgment vehicle, and based on the road network model and path algorithm, obtains the first driving path, the first theoretical path and multiple first theoretical fees, and then obtains the first judgment result, the second judgment result and the third judgment result. This multi-angle judgment can improve the accuracy of the final toll evasion judgment, and thus improve the accuracy of the audit of vehicle toll evasion.

[0156] Step S105: Calculate the highway toll payable for each vehicle in the second toll evasion vehicle feature set based on the vehicle traffic data, and conduct a vehicle toll evasion audit based on the toll payable.

[0157] In certain embodiments of the present application, the calculating, based on the vehicle traffic data, the highway toll payable for each vehicle in the second toll evasion vehicle feature set specifically includes:

[0158] Sequentially selecting each vehicle in the second toll evasion vehicle feature set as the current toll-billing vehicle, and obtaining the toll payable for the current toll-billing vehicle according to a preset toll evasion billing method, until all vehicles in the second toll evasion vehicle feature set have been selected;

[0159] The method for charging evasion fees is as follows:

[0160] Determine the entry and exit data of the toll station corresponding to the current billing vehicle based on multiple vehicle travel times of the current billing vehicle;

[0161] If both the entry data and the exit data are not empty, the highway toll payable for the current billing vehicle is obtained based on the entry data and the exit data;

[0162] If either the entry data or the exit data is empty, the second theoretical path of the current billing vehicle is obtained based on the vehicle passage data of the current billing vehicle, and the highway toll that should be paid by the current billing vehicle is obtained based on the second theoretical path.

[0163] In certain embodiments of the present application, the highway toll payable for the current billing vehicle is obtained based on the entry data and the exit data, specifically:

[0164] Obtaining the actual highway toll paid by the current billing vehicle based on the exit data;

[0165] Obtaining the highway toll payable for the current billing vehicle based on the entry data and the exit data;

[0166] The toll payable for the current vehicle is obtained based on the actually paid toll and the payable toll.

[0167] In certain embodiments of the present application, the second theoretical path of the current billing vehicle is obtained based on the vehicle passage data of the current billing vehicle, specifically:

[0168] According to the vehicle passage data of the current billing vehicle, a plurality of fourth toll stations corresponding to the current billing vehicle are obtained;

[0169] Based on the road network model and the vehicle traffic data of the current billing vehicle, the multiple fourth toll stations are connected to obtain the second theoretical path of the current billing vehicle.

[0170] In certain embodiments of the present application, the highway toll payable for the current charging vehicle is obtained according to the second theoretical path, specifically:

[0171] According to the second theoretical path, the highway toll payable by the current billing vehicle is obtained;

[0172] The payable highway toll is used as the highway toll payable for the current billing vehicle.

[0173] This application first selects the current billing vehicle, and then determines the corresponding toll station's entry and exit data based on the multiple vehicle travel times of the current billing vehicle, and calculates the highway toll that the current billing vehicle should pay based on the null value of the entry and exit data. This can improve the accuracy of calculating the highway toll that should be paid for vehicles that evade fees, and thereby improve the accuracy of auditing vehicles that evade fees.

[0174] Compared with the existing technology, the present application first extracts features from vehicle traffic data to obtain a passing vehicle feature set, and then divides the passing vehicle feature set to obtain a passing vehicle feature set and a road network vehicle feature set, and then obtains a first toll-evading vehicle feature set based on a feature difference scoring method. It can improve the recognition accuracy of similar vehicles and reduce the probability of misjudgment by comparing the feature differences between vehicles passing through toll stations and vehicles passing through expressway sections, and then screen the first toll-evading vehicle feature set based on the road network model to obtain a second toll-evading vehicle feature set. It can screen out some normal passing vehicles that are not toll-evading based on the road network model, and then improve the accuracy of vehicle toll-evading audit when calculating the highway tolls to be paid for each vehicle in the second toll-evading vehicle feature set for vehicle toll-evading audit.

[0175] Corresponding to the above method, see Figure 2 The embodiment of the present application provides a vehicle fee evasion audit device, including a vehicle feature extraction module 210, a feature set segmentation module 220, a first fee evasion set determination module 230, a second fee evasion set determination module 240, and a vehicle fee supplement calculation module 250;

[0176] The vehicle feature extraction module 210 is configured to obtain vehicle traffic data from multiple toll booths and multiple highway sections, and perform feature extraction on the vehicle traffic data to obtain a passing vehicle feature set; wherein the multiple highway sections are located between the multiple toll booths; the vehicle traffic data includes vehicle traffic time, vehicle identification features, and multiple frames of vehicle traffic images;

[0177] The feature set segmentation module 220 is used to segment the passing vehicle feature set based on the preset passing feature markers to obtain a passing vehicle feature set and a road network vehicle feature set;

[0178] The first toll evasion set determination module 230 is configured to obtain a first toll evasion vehicle feature set based on the transit vehicle feature set and the road network vehicle feature set based on a preset feature difference scoring method;

[0179] The second fee evasion set determination module 240 is configured to filter the first fee evasion vehicle feature set based on a preset road network model to obtain a second fee evasion vehicle feature set;

[0180] The vehicle toll calculation module 250 is used to calculate the highway toll that should be paid for each vehicle in the second toll evasion vehicle feature set based on the vehicle passage data, so as to conduct vehicle toll evasion audit based on the highway toll that should be paid.

[0181] In certain embodiments of the present application, the vehicle feature extraction module 210 includes an image segmentation submodule 211 , an image normalization submodule 212 , a feature specification submodule 213 , and a feature acquisition submodule 214 ;

[0182] The image segmentation submodule 211 is configured to segment a plurality of vehicle traffic images of each vehicle in the vehicle traffic data according to a deep learning-based image segmentation method to obtain vehicle color features, vehicle model features, and vehicle damage features of each vehicle;

[0183] The image normalization submodule 212 is used to normalize and combine the vehicle color feature, vehicle model feature, and vehicle damage feature of each vehicle to obtain a first comprehensive feature of each vehicle;

[0184] The feature standardization submodule 213 is used to standardize the first comprehensive feature of each vehicle based on a preset standard to obtain a second comprehensive feature of each vehicle;

[0185] The feature acquisition submodule 214 is configured to obtain a passing vehicle feature set based on the second comprehensive feature of each vehicle, the vehicle identification feature, and the vehicle passing time.

[0186] In certain embodiments of the present application, the first evasion fee set determination module 230 includes a first evasion fee set determination submodule 231;

[0187] The first fee evasion set determination submodule 231 is configured to sequentially select each vehicle in the transit vehicle feature set as the current first vehicle, obtain an audit result of the current first vehicle according to a preset vehicle fee evasion audit method, and obtain the first fee evasion vehicle feature set based on the audit result until all vehicles in the transit vehicle feature set have been selected.

[0188] The method for auditing vehicle fee evasion is as follows:

[0189] Calculating a feature difference score between the current first vehicle and each vehicle in the road network vehicle feature set based on a preset feature difference scoring method;

[0190] If any one of the multiple feature difference scores of the current first vehicle reaches a preset threshold, the license plate numbers of the current first vehicle and the current second vehicle are obtained based on several frames of vehicle traffic images of the current first vehicle and several frames of vehicle traffic images of the current second vehicle; wherein the current second vehicle is a vehicle in the road network vehicle feature set corresponding to the feature difference score that reaches the preset threshold;

[0191] If the license plate number of the current first vehicle is different from the license plate number of the current second vehicle, the vehicle registration information of the current first vehicle is obtained based on the vehicle identification characteristics of the current first vehicle, and the audit result of the current first vehicle is obtained based on the vehicle registration information.

[0192] In certain embodiments of the present application, the feature difference scoring method is specifically:

[0193] Sequentially selecting each vehicle in the road network vehicle feature set as a current scoring vehicle, and calculating a feature difference score between the current first vehicle and the current scoring vehicle according to a preset first scoring method, until all vehicles in the road network vehicle feature set have been selected;

[0194] The first scoring method is specifically:

[0195] Based on a similarity algorithm, multiple similarities between the current first vehicle and the currently rated vehicle are calculated; wherein the multiple similarities are based on vehicle color features, vehicle model features, vehicle damage features, and vehicle identification features of the current first vehicle and the currently rated vehicle;

[0196] A feature difference score between the current first vehicle and the current scoring vehicle is obtained based on the multiple similarities.

[0197] In certain embodiments of the present application, the second evasion fee set determination module 240 includes a second evasion fee set determination submodule 241;

[0198] The second fee evasion set determination submodule 241 is configured to sequentially select each vehicle in the first fee evasion vehicle feature set as a first identification vehicle, and obtain a fee evasion identification result for the first identification vehicle according to a preset vehicle fee evasion identification method, until all vehicles in the first fee evasion vehicle feature set have been selected;

[0199] The method for determining whether a vehicle has evaded or missed charges is as follows:

[0200] Acquire multiple toll station passage data corresponding to the first identified vehicle; wherein the toll station passage data includes toll station location data, vehicle passing time, and vehicle toll fees;

[0201] Based on a preset road network model, according to the toll station location data and the vehicle's transit time, a first driving path of the first identified vehicle is obtained;

[0202] Based on the road network model and the preset path algorithm, and according to the toll station location data, a first theoretical path and a plurality of first theoretical fees for the first identified vehicle are obtained;

[0203] Obtaining a first vehicle identification result based on the first driving path and the first theoretical path;

[0204] Obtaining a second discrimination result of the first discrimination vehicle based on the vehicle toll and the plurality of first theoretical fees;

[0205] Obtaining a third discrimination result of the first discrimination vehicle according to the vehicle passing time;

[0206] According to the first discrimination result, the second discrimination result and the third discrimination result, a discrimination result of the first discrimination vehicle for toll evasion is obtained.

[0207] In certain embodiments of the present application, the vehicle fee calculation module 250 includes a vehicle fee calculation submodule 251;

[0208] The vehicle toll calculation submodule 251 is configured to sequentially select each vehicle in the second toll evasion vehicle feature set as the current toll-charging vehicle, and calculate the toll payable for the current toll-charging vehicle according to a preset toll evasion billing method, until all vehicles in the second toll evasion vehicle feature set have been selected;

[0209] The method for charging evasion fees is as follows:

[0210] Determine the entry and exit data of the toll station corresponding to the current billing vehicle based on multiple vehicle travel times of the current billing vehicle;

[0211] If both the entry data and the exit data are not empty, the highway toll payable for the current billing vehicle is obtained based on the entry data and the exit data;

[0212] If either the entry data or the exit data is empty, the second theoretical path of the current billing vehicle is obtained based on the vehicle passage data of the current billing vehicle, and the highway toll that should be paid by the current billing vehicle is obtained based on the second theoretical path.

[0213] Compared with the existing technology, the present application first extracts features from vehicle traffic data to obtain a passing vehicle feature set, and then divides the passing vehicle feature set to obtain a passing vehicle feature set and a road network vehicle feature set, and then obtains a first toll-evading vehicle feature set based on a feature difference scoring method. It can improve the recognition accuracy of similar vehicles and reduce the probability of misjudgment by comparing the feature differences between vehicles passing through toll stations and vehicles passing through expressway sections, and then screen the first toll-evading vehicle feature set based on the road network model to obtain a second toll-evading vehicle feature set. It can screen out some normal passing vehicles that are not toll-evading based on the road network model, and then improve the accuracy of vehicle toll-evading audit when calculating the highway tolls to be paid for each vehicle in the second toll-evading vehicle feature set for vehicle toll-evading audit.

[0214] It should be understood that the device provided in the embodiment of the present application corresponds to the aforementioned method, and the vehicle fee evasion and auditing device provided in the embodiment of the present application can implement the vehicle fee evasion and auditing method provided in any embodiment of the present application.

[0215] Adaptively, the embodiments of the present application further provide a computer device and a computer-readable storage medium.

[0216] The computer device comprises: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor;

[0217] Among them, when the processor executes the computer program, a vehicle fee evasion and fee verification method of the present application is implemented.

[0218] The computer-readable storage medium stores a plurality of instructions, which are suitable for loading by a processor to execute a vehicle fee evasion and auditing method of the present application.

[0219] The above description is a partial embodiment of the present application, which further describes the purpose, technical solutions, and beneficial effects of the present application in detail. It should be understood that the above description of the partial embodiment of the present application is not to be construed as limiting the present application. In particular, it is pointed out that for those skilled in the art, any changes, modifications, equivalent substitutions, and variations made within the spirit and principles of the present application should be included within the scope of protection of the present application.

Claims

1. A method for auditing vehicle fee evasion, characterized in that: include: Acquiring vehicle traffic data from multiple toll booths and multiple highway sections, and performing feature extraction on the vehicle traffic data to obtain a feature set of passing vehicles; wherein the multiple highway sections are located between the multiple toll booths; the vehicle traffic data includes vehicle traffic time, vehicle identification features, and multiple frames of vehicle traffic images; Based on the preset passing feature markers, the passing vehicle feature set is segmented to obtain a passing vehicle feature set and a road network vehicle feature set; Based on a preset feature difference scoring method, a first toll evasion vehicle feature set is obtained according to the transit vehicle feature set and the road network vehicle feature set; Based on a preset road network model, the first feature set of vehicles that evade tolls is screened to obtain a second feature set of vehicles that evade tolls; Calculating the toll payable for each vehicle in the second toll evasion vehicle feature set based on the vehicle traffic data, and conducting a toll evasion audit based on the toll payable; The method of screening the first toll-evading vehicle feature set based on a preset road network model to obtain a second toll-evading vehicle feature set specifically includes: Sequentially selecting each vehicle in the first vehicle feature set as a first identification vehicle, and obtaining an identification result of the first identification vehicle according to a preset vehicle identification method for evading tolls, until all vehicles in the first vehicle feature set have been selected; Among them, the method for identifying vehicles that evade tolls is specifically as follows: obtaining multiple toll station passage data corresponding to a first identification vehicle; wherein, the toll station passage data includes toll station location data, vehicle passing time and vehicle toll fees; based on a preset road network model, according to the toll station location data and the vehicle passing time, obtaining a first driving path of the first identification vehicle; based on the road network model and a preset path algorithm, according to the toll station location data, obtaining a first theoretical path and multiple first theoretical fees of the first identification vehicle; according to the first driving path and the first theoretical path, obtaining a first identification result of the first identification vehicle; according to the vehicle toll fees and the multiple first theoretical fees, obtaining a second identification result of the first identification vehicle; according to the vehicle passing time, obtaining a third identification result of the first identification vehicle; according to the first identification result, the second identification result and the third identification result, obtaining a toll evasion identification result of the first identification vehicle.

2. A vehicle fee evasion and auditing method according to claim 1, characterized in that: The feature extraction of the vehicle traffic data to obtain a feature set of passing vehicles specifically includes: Segmenting a plurality of vehicle traffic images of each vehicle in the vehicle traffic data according to a deep learning-based image segmentation method to obtain vehicle color features, vehicle model features, and vehicle damage features of each vehicle; Normalize and concatenate the vehicle color feature, vehicle model feature, and vehicle damage feature of each vehicle to obtain the first comprehensive feature of each vehicle; Based on a preset standard, the first comprehensive feature of each vehicle is normalized to obtain a second comprehensive feature of each vehicle; A passing vehicle feature set is obtained based on the second comprehensive feature of each vehicle, the vehicle identification feature and the vehicle passing time.

3. A vehicle fee evasion and auditing method according to claim 2, characterized in that: The method based on the preset feature difference scoring method obtains a first feature set of toll evasion vehicles according to the feature set of transit vehicles and the feature set of road network vehicles, specifically including: Sequentially selecting each vehicle in the transit vehicle feature set as the current first vehicle, and obtaining an audit result of the current first vehicle according to a preset vehicle fee evasion audit method, until all vehicles in the transit vehicle feature set are selected, and obtaining a first fee evasion vehicle feature set according to the audit result; The method for auditing vehicle fee evasion is as follows: Calculating a feature difference score between the current first vehicle and each vehicle in the road network vehicle feature set based on a preset feature difference scoring method; If any one of the multiple feature difference scores of the current first vehicle reaches a preset threshold, the license plate numbers of the current first vehicle and the current second vehicle are obtained based on several frames of vehicle traffic images of the current first vehicle and several frames of vehicle traffic images of the current second vehicle; wherein the current second vehicle is a vehicle in the road network vehicle feature set corresponding to the feature difference score that reaches the preset threshold; If the license plate number of the current first vehicle is different from the license plate number of the current second vehicle, the vehicle registration information of the current first vehicle is obtained based on the vehicle identification characteristics of the current first vehicle, and the audit result of the current first vehicle is obtained based on the vehicle registration information.

4. A vehicle fee evasion and auditing method according to claim 3, characterized in that: The feature difference scoring method is specifically as follows: Sequentially selecting each vehicle in the road network vehicle feature set as a current scoring vehicle, and calculating a feature difference score between the current first vehicle and the current scoring vehicle according to a preset first scoring method, until all vehicles in the road network vehicle feature set have been selected; The first scoring method is specifically: Based on a similarity algorithm, multiple similarities between the current first vehicle and the currently rated vehicle are calculated; wherein the multiple similarities are based on vehicle color features, vehicle model features, vehicle damage features, and vehicle identification features of the current first vehicle and the currently rated vehicle; A feature difference score between the current first vehicle and the current scoring vehicle is obtained based on the multiple similarities.

5. The vehicle fee evasion and fee auditing method according to claim 1, characterized in that: The calculating, based on the vehicle traffic data, the highway toll payable for each vehicle in the second toll evasion vehicle feature set specifically includes: Sequentially selecting each vehicle in the second toll evasion vehicle feature set as the current toll-billing vehicle, and obtaining the toll payable for the current toll-billing vehicle according to a preset toll evasion billing method, until all vehicles in the second toll evasion vehicle feature set have been selected; The method for charging evasion fees is as follows: Determine the entry and exit data of the toll station corresponding to the current billing vehicle based on multiple vehicle travel times of the current billing vehicle; If both the entry data and the exit data are not empty, the highway toll payable for the current billing vehicle is obtained based on the entry data and the exit data; If either the entry data or the exit data is empty, the second theoretical path of the current billing vehicle is obtained based on the vehicle passage data of the current billing vehicle, and the highway toll that should be paid by the current billing vehicle is obtained based on the second theoretical path.

6. A vehicle fee evasion and auditing device, characterized in that: It includes a vehicle feature extraction module, a feature set segmentation module, a first evasion fee set determination module, a second evasion fee set determination module and a vehicle fee supplement calculation module; The vehicle feature extraction module is configured to obtain vehicle traffic data from multiple toll booths and multiple highway sections, and perform feature extraction on the vehicle traffic data to obtain a traffic feature set; wherein the multiple highway sections are located between the multiple toll booths; the vehicle traffic data includes vehicle traffic time, vehicle identification features, and multiple frames of vehicle traffic images; The feature set segmentation module is used to segment the passing vehicle feature set based on the preset passing feature markers to obtain the passing vehicle feature set and the road network vehicle feature set; The first toll evasion set determination module is configured to obtain a first toll evasion vehicle feature set based on the transit vehicle feature set and the road network vehicle feature set based on a preset feature difference scoring method; The second fee evasion set determination module is configured to filter the first fee evasion vehicle feature set based on a preset road network model to obtain a second fee evasion vehicle feature set; The vehicle fee calculation module is used to calculate the highway fee that should be paid for each vehicle in the second feature set of vehicles that evade tolls based on the vehicle traffic data, so as to conduct vehicle fee evasion audits based on the highway fee that should be paid; The second fee evasion set determination module includes a second fee evasion set determination submodule; The second fee evasion set determination submodule is configured to sequentially select each vehicle in the first fee evasion vehicle feature set as a first discrimination vehicle, and obtain a fee evasion discrimination result for the first discrimination vehicle according to a preset vehicle fee evasion discrimination method, until all vehicles in the first fee evasion vehicle feature set have been selected; Among them, the method for identifying vehicles that evade tolls is specifically as follows: obtaining multiple toll station passage data corresponding to a first identification vehicle; wherein, the toll station passage data includes toll station location data, vehicle passing time and vehicle toll fees; based on a preset road network model, according to the toll station location data and the vehicle passing time, obtaining a first driving path of the first identification vehicle; based on the road network model and a preset path algorithm, according to the toll station location data, obtaining a first theoretical path and multiple first theoretical fees of the first identification vehicle; according to the first driving path and the first theoretical path, obtaining a first identification result of the first identification vehicle; according to the vehicle toll fees and the multiple first theoretical fees, obtaining a second identification result of the first identification vehicle; according to the vehicle passing time, obtaining a third identification result of the first identification vehicle; according to the first identification result, the second identification result and the third identification result, obtaining a toll evasion identification result of the first identification vehicle.

7. The vehicle fee evasion and fee auditing device according to claim 6, characterized in that: The vehicle feature extraction module includes an image segmentation submodule, an image normalization submodule, a feature specification submodule and a feature acquisition submodule; The image segmentation submodule is used to segment several frames of vehicle traffic images of each vehicle in the vehicle traffic data according to an image segmentation method based on deep learning, so as to obtain vehicle color features, vehicle model features and vehicle damage features of each vehicle; The image normalization submodule is used to normalize and splice the vehicle color features, vehicle model features, and vehicle damage features of each vehicle to obtain a first comprehensive feature of each vehicle; The feature normalization submodule is configured to normalize the first comprehensive feature of each vehicle based on a preset standard to obtain a second comprehensive feature of each vehicle; The feature acquisition submodule is used to obtain a passing vehicle feature set based on the second comprehensive feature of each vehicle, the vehicle identification feature and the vehicle passing time.

8. The vehicle fee evasion and fee auditing device according to claim 7, characterized in that: The first fee evasion and leakage set determination module includes a first fee evasion and leakage set determination submodule; The first fee evasion set determination submodule is configured to sequentially select each vehicle in the transit vehicle feature set as the current first vehicle, obtain an audit result of the current first vehicle according to a preset vehicle fee evasion audit method, and obtain the first fee evasion vehicle feature set based on the audit result until all vehicles in the transit vehicle feature set have been selected. The method for auditing vehicle fee evasion is as follows: Calculating a feature difference score between the current first vehicle and each vehicle in the road network vehicle feature set based on a preset feature difference scoring method; If any one of the multiple feature difference scores of the current first vehicle reaches a preset threshold, the license plate numbers of the current first vehicle and the current second vehicle are obtained based on several frames of vehicle traffic images of the current first vehicle and several frames of vehicle traffic images of the current second vehicle; wherein the current second vehicle is a vehicle in the road network vehicle feature set corresponding to the feature difference score that reaches the preset threshold; If the license plate number of the current first vehicle is different from the license plate number of the current second vehicle, the vehicle registration information of the current first vehicle is obtained based on the vehicle identification characteristics of the current first vehicle, and the audit result of the current first vehicle is obtained based on the vehicle registration information.

9. The vehicle fee evasion and fee auditing device according to claim 8, characterized in that: The feature difference scoring method is specifically as follows: Sequentially selecting each vehicle in the road network vehicle feature set as a current scoring vehicle, and calculating a feature difference score between the current first vehicle and the current scoring vehicle according to a preset first scoring method, until all vehicles in the road network vehicle feature set have been selected; The first scoring method is specifically: Based on a similarity algorithm, multiple similarities between the current first vehicle and the currently rated vehicle are calculated; wherein the multiple similarities are based on vehicle color features, vehicle model features, vehicle damage features, and vehicle identification features of the current first vehicle and the currently rated vehicle; A feature difference score between the current first vehicle and the current scoring vehicle is obtained based on the multiple similarities.

10. The vehicle fee evasion and fee auditing device according to claim 6, characterized in that: The vehicle fee calculation module includes a vehicle fee calculation submodule; The vehicle toll calculation submodule is configured to sequentially select each vehicle in the second toll evasion vehicle feature set as the current toll-charging vehicle, and obtain the toll payable for the current toll-charging vehicle according to a preset toll evasion billing method, until all vehicles in the second toll evasion vehicle feature set have been selected; The method for charging evasion fees is as follows: Determine the entry and exit data of the toll station corresponding to the current billing vehicle based on multiple vehicle travel times of the current billing vehicle; If both the entry data and the exit data are not empty, the highway toll payable for the current billing vehicle is obtained based on the entry data and the exit data; If either the entry data or the exit data is empty, the second theoretical path of the current billing vehicle is obtained based on the vehicle passage data of the current billing vehicle, and the highway toll that should be paid by the current billing vehicle is obtained based on the second theoretical path.

Images