High-speed audit system based on path algorithm

Abstract

The present invention discloses a high-speed audit system based on a path algorithm, relates to the technical field of high-speed audit, and solves the technical problem that the existing scheme does not take into account the scenario of cross-provincial travel, has high implementation cost and operation complexity, and cannot accurately and efficiently identify the technical problem of vehicles of long-distance buying and short-distance buying; the system comprises a processor and a data acquisition module connected to the processor in communication; the present invention generates a gantry flow model by integrating the gantry transaction flow and the gantry plate recognition flow, and screens the provincial boundary gantry flow model and the border gantry flow model according to the gantry flow model; effectively merges the high-speed cross-provincial travel data, and improves the recognition efficiency and accuracy of long-distance buying and short-distance buying; the present invention combines manual verification with an artificial intelligence model, combines continuous verification data feedback during business use, continuously optimizes the model, can improve recognition accuracy, and at the same time improves audit efficiency.

Description

Technical Field

[0001] The present invention belongs to the technical field of high-speed auditing, and in particular to a high-speed auditing system based on a path algorithm. Background Art

[0002] As an important business link in the highway network toll collection business, highway auditing is an important means to safeguard the legitimate interests of business entities and ensure the fairness of toll collection business.

[0003] Relying on gantry technology, the driving path of passing vehicles on the highway can be accurately fitted, so that tolls can be collected according to the actual driving path; but in the actual collection process, it was found that running long and buying short is still the most stubborn reason for the loss of tolls; the existing solution to filter the data of running long and buying short is to associate and filter the gantry transaction records, card recognition flow, and inlet and outlet flow according to the rules, and according to the external data rules reflected by running long and buying short, the corresponding data can be found.

[0004] The existing solutions do not take into account the scenario of cross-provincial travel, and the implementation cost and operation complexity are relatively high. Therefore, there is an urgent need for a high-level audit system that is simple to operate, low-cost, and can accurately identify long-term and short-term scams. Summary of the invention

[0005] The present invention provides a high-speed audit system based on a path algorithm, which is used to solve the technical problems that the existing solutions do not take into account the scenario of cross-provincial travel, have high implementation costs and operation complexity, and cannot accurately and efficiently identify vehicles that run long and buy short. The present invention solves the above problems by merging gantry transaction and license plate recognition data, removing duplicate boundary flow data, and assisting in auditing through an artificial intelligence model.

[0006] The object of the present invention can be achieved by the following technical solution: a high-speed audit system based on a path algorithm, comprising a processor, and a data acquisition module in communication with the processor;

[0007] The audit support data is collected through the data collection module, and the audit support data is preprocessed; wherein the audit support data includes basic data and flow data;

[0008] Integrate the gantry transaction flow and the gantry brand recognition flow to generate a gantry flow model, and select the provincial boundary gantry flow model and the border gantry flow model based on the gantry flow model;

[0009] Integrate the lane entrance flow, provincial boundary gantry transaction flow, and border gantry transaction flow to generate an entrance flow model;

[0010] The lane exit flow is associated with the entrance flow model, and the exit transactions are extracted from the provincial boundary gantry flow model and the gantry sign recognition flow, and then merged to generate the exit flow model;

[0011] The exit flow model is used as a baseline for traversal, and the door frame license plate identifications queried are sorted and deduplicated according to time, and a license plate identification path charging unit combination is generated;

[0012] The export flow model is used as a baseline for traversal, and after the queried gantry transactions are sorted and deduplicated according to time, a transaction path charging unit combination is generated, and a high-frequency path is obtained according to the transaction path charging unit combination;

[0013] The similarity of the transaction path charging unit combination and the card recognition path charging unit combination is calculated and the audit fee amount is obtained; the audit suspicion is calculated based on the similarity threshold and the audit fee amount.

[0014] Preferably, the provincial boundary gantry flow model and the border gantry flow model are screened out by associating the gantry flow model with the gantry normative definition data.

[0015] Preferably, the basic data includes normative definition data of toll stations and gantries and normative definition data of toll units; the flow data includes lane entrance flow, lane exit flow, gantry transaction flow and gantry plate recognition flow.

[0016] Preferably, the audit suspicion is obtained through an artificial intelligence model; wherein the artificial intelligence model includes a neural network model and a decision tree model.

[0017] Preferably, vehicle information data is extracted from the flow data, and the vehicle information data is integrated and written into a vehicle model library; the vehicle model library counts the vehicle information data corresponding to each license plate that appears.

[0018] Compared with the prior art, the present invention has the following beneficial effects:

[0019] 1. The present invention integrates the gantry transaction flow and the gantry brand recognition flow to generate a gantry flow model, and screens the provincial boundary gantry flow model and the border gantry flow model according to the gantry flow model; it effectively merges the high-speed inter-provincial traffic data and improves the recognition efficiency and accuracy of running long and buying short.

[0020] 2. The present invention combines manual verification with artificial intelligence models, and continuously optimizes the model based on continuous verification data feedback during business use, which can improve recognition accuracy and audit efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0022] Figure 1 It is a schematic diagram of the workflow of the present invention. DETAILED DESCRIPTION

[0023] The technical solution of the present application will be described clearly and completely in conjunction with the embodiments below. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present application.

[0024] The terms used herein are used to describe embodiments and are not intended to restrict and / or limit the present disclosure; it should be noted that the singular forms "a", "an", and "the" also include plural forms unless the context clearly indicates otherwise; and, although terms "first", "second", etc. may be used herein to describe various elements, the elements are not limited by these terms, which are only used to distinguish one element from another.

[0025] See also Figure 1 For those who travel long and buy short, the scenario of cross-provincial traffic is often not taken into consideration. Generally, the vehicle trip within the province from the toll station in the province to the toll station in the province is selected as the screening scenario. This excludes a considerable proportion of cross-provincial toll evasion in terms of both amount and quantity. In addition, although the toll stations at provincial boundaries have been abolished, there may not be only one highway operator in each province. In the case of only a single road network data, how to audit the road network traffic across operators is also a thorny issue. In response to the above problems, the present invention provides a high-speed audit system based on a path algorithm, including a processor and a data acquisition module that communicates with and / or is electrically connected to the processor; the processor is used to complete data processing, and the data acquisition module is used to complete data collection.

[0026] In the embodiment of the present invention, the audit support data is collected by the data collection module, and the audit support data is preprocessed; the audit support data includes basic data and flow data.

[0027] In a specific embodiment, the basic data includes normative definition data of toll stations and gantries and normative definition data of toll units; normative definition data of toll stations and gantries such as names, codes, Hex codes and other data; normative definition data of toll units include but are not limited to names, Hex codes, and provincial boundary identification (to distinguish between entry and exit provincial boundaries).

[0028] In a specific embodiment, the flow data includes lane entrance flow, lane exit flow, gantry transaction flow and gantry license plate recognition flow; the lane entrance flow includes the entrance station Hex, entrance time, pass sign, license plate, vehicle model, vehicle type and other data; the lane exit flow includes but is not limited to the exit station Hex, exit time, pass sign, license plate, vehicle model, vehicle type, entrance station Hex, entrance time, entrance license plate, entrance vehicle model, entrance vehicle type, toll payable, actual toll amount, passage medium type, billing type and other data; the gantry transaction flow includes but is not limited to the gantry Hex code, passage time, pass sign, license plate, vehicle model, vehicle type and other data; the gantry license plate recognition flow includes but is not limited to the gantry Hex code, passage time, license plate and other data.

[0029] It is worth noting that in this embodiment, the road sections under the jurisdiction of the road network unit are sorted out, and the gantries located at the boundaries of the range are used as boundary gantries. This scheme believes that provincial boundary gantries and boundary gantries can be equivalent to toll stations in cross-provincial and cross-border travel scenarios.

[0030] In an embodiment of the present invention, the gantry transaction flow and the gantry plate recognition flow are integrated to generate a gantry flow model, and the provincial boundary gantry flow model and the border gantry flow model are screened according to the gantry flow model; in view of the fact that the gantry transaction and the gantry flow do not have a unique binding relationship, and both may have data duplication and data loss, the present application starts from the time interval, merges the data within the threshold of the time interval between the two and regards them as the same gantry passage, and retains the unmatched data of the two, so as to restore the vehicle's passage trajectory as much as possible.

[0031] In a specific embodiment, the gantry flow model is associated with the gantry normative definition data, and the provincial boundary gantry flow model and the border gantry flow model are screened out. Since the area under the jurisdiction of the road network center may be a concave polygon, here, by judging whether the path contains continuous outbound and inbound gantries and their relative positions, the scene of actually including several cross-border passes in one pass is merged. The result of this step is the provincial boundary and border gantry data.

[0032] In the embodiment of the present invention, the lane entrance flow, the provincial boundary gantry transaction flow, and the border gantry transaction flow are integrated to generate an entrance flow model.

[0033] In an embodiment of the present invention, the lane exit flow is associated with the entrance flow model, the exit transaction is extracted from the provincial boundary gantry flow model and the gantry sign recognition flow, and the exit flow model is generated after merging; it includes two steps: 1) the lane exit flow is associated with the entrance flow model, and the corresponding entrance information is supplemented; 2) the exit transaction is extracted from the provincial boundary gantry transaction and the provincial boundary gantry sign recognition. Note that the merging is required, that is, the gantry sign recognition associated with the gantry transaction is regarded as one transaction.

[0034] For export transactions generated by gantry data at provincial borders, entry data needs to be supplemented. The principle of supplementation is to first make a strong association through the pass mark, and then supplement the entry data if it is associated; for those that are not associated, use "license plate + nearest time" to match the entry / exit, and supplement all entry data if the entry station is successfully matched. If other methods successfully match, only the entry Hex, entry type and entry time are supplemented.

[0035] In the embodiment of the present invention, the export flow model is used as a baseline for traversal, and the queried door frame license plate identifications are sorted by time to remove duplicates, and then a license plate identification path charging unit combination is generated.

[0036] In a specific embodiment, the generated exit flow model is used as a baseline for traversal, and the nearest passage cut-off time is supplemented. The supplementation principle is to use "license plate + nearest time" to match the time of exit / provincial boundary station sign recognition, and the traversed transactions are queried for gantry transaction records according to the pass identification. The queried gantry plate recognitions are sorted by time and de-duplicated to form a plate recognition path charging unit combination. In this process, multiple cards in one vehicle can be identified based on the gantry code, and the similarity rate of the transaction path charging unit combination and the plate recognition path charging unit combination is calculated.

[0037] In the embodiment of the present invention, the export flow model is used as a baseline for traversal, and the queried gantry transactions are sorted by time to remove duplicates, and a transaction path charging unit combination is generated, and a high-frequency path is obtained based on the transaction path charging unit combination.

[0038] In a specific embodiment, the generated export flow model is used as the baseline for traversal. If the data cardinality is large, only transactions with access identification are traversed. For export transactions without access identification, the license plate and transaction can be used to query the gantry license plate recognition records according to the start and end time. The queried gantry transactions are sorted by time and deduplicated to form a transaction path charging unit combination. The MD5 digest of the transaction path charging unit combination is calculated, and a key-value pair of MD5 digest-charging unit combination is formed. A "one-to-many" mapping relationship is further formed with "entrance + exit" as the key and MD5 digest as the value. The data of the previous step is traversed, aggregated and counted, and the number of occurrences of each charging unit combination (MD5 digest) is calculated. The data of the previous step is traversed again to calculate the high-frequency path (MD5 digest).

[0039] In the embodiment of the present invention, the similarity of the transaction path charging unit combination and the card recognition path charging unit combination is calculated and the audit fee amount is obtained; the audit suspicion is calculated based on the similarity threshold and the audit fee amount.

[0040] In a specific embodiment, the similarity threshold is marked as XSY, the actual similarity is marked as SCL, and the audit fee amount is marked as JSE; the audit suspicion JXD is obtained by the formula JXD=ɑ×|XSY-SCL|×|JSE-SEY|; wherein SEY is the fee amount threshold, ɑ is the proportional coefficient, and ɑ is a real number greater than 0; it can be understood that the greater the difference between the actual similarity and the similarity and / or the greater the difference between the audit fee amount and the fee amount threshold, the higher the audit suspicion.

[0041] In a specific embodiment, a similarity calculation is performed on the transaction path charging unit combination and the license plate recognition path charging unit combination, the overlapping path portion of the gantry transaction and the license plate recognition is then similarly calculated with the high-frequency path, the audit fee amount is calculated for the overlapping path portion, a U / J conversion judgment is performed on the overlapping path portion, and the data of a single pass is supplemented. Based on the single pass record, the audit suspicion records are filtered out according to the set similarity threshold and audit fee amount, and the audit suspicion degree is calculated.

[0042] In some optional embodiments, the path fitting display function is used to assist manual judgment on the rationality of the gantry transaction path and the gantry brand recognition path.

[0043] In some optional embodiments, vehicle evasion behavior is confirmed through ministerial and provincial platforms, data feedback is provided, parameter threshold settings are adjusted, and used as a data set for machine learning; decision tree and neural network machine learning models are trained to achieve intelligent evasion identification.

[0044] In some optional embodiments, vehicle information data such as license plate, vehicle model, vehicle type, number of axles, and maximum number of passengers are extracted from the flow data and written into the vehicle model library according to a certain logic. The vehicle model library records the vehicle information data corresponding to each license plate appearing in the flow and counts them separately. Whenever there is an information conflict, it is necessary to use frequency logic to determine whether the data is credible; the highest confidence data of each vehicle in the vehicle model library is used as the audit model and participates in the calculation of the path audit amount.

[0045] In some optional embodiments, behaviors such as multiple cards on one vehicle and shielded OBU can be discovered through gantry transaction data and gantry license plate recognition data, and behaviors such as vehicle type inconsistency (large vehicle with small label) can be discovered through inconsistent axle number at the entrance and exit.

[0046] The data in the above formula are all calculated by removing the dimensions and taking their numerical values. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters and preset thresholds in the formula are set by technical personnel in this field according to actual conditions or obtained by simulating a large amount of data.

[0047] How this application works:

[0048] Audit support data is collected through the data collection module, and data preprocessing is performed on the audit support data; the gantry transaction flow and the gantry brand recognition flow are integrated to generate a gantry flow model, and the provincial boundary gantry flow model and the border gantry flow model are screened according to the gantry flow model.

[0049] Integrate the lane entrance flow, provincial boundary gantry transaction flow, and border gantry transaction flow to generate an entrance flow model; associate the lane exit flow with the entrance flow model, extract exit transactions from the provincial boundary gantry flow model and the gantry plate recognition flow, and merge them to generate an exit flow model.

[0050] The export flow model is used as a baseline for traversal, and the queried gantry license plate identifications are sorted by time and de-duplicated, and a license plate identification path charging unit combination is generated; the export flow model is used as a baseline for traversal, and the queried gantry transactions are sorted by time and de-duplicated, and a transaction path charging unit combination is generated, and a high-frequency path is obtained according to the transaction path charging unit combination.

[0051] The similarity of the transaction path charging unit combination and the card recognition path charging unit combination is calculated and the audit fee amount is obtained; the audit suspicion is calculated based on the similarity threshold and the audit fee amount.

[0052] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0053] The above contents are merely examples and explanations of the structure of the present application. The technical personnel in the relevant technical field may make various modifications or additions to the specific embodiments described or replace them in a similar manner. As long as they do not deviate from the structure of the application or exceed the scope defined by the claims, they should all fall within the scope of protection of the present application.

Claims

1. A high-speed audit system based on a path algorithm, comprising a processor and a data acquisition module connected to the processor in communication, characterized in that: include: The audit support data is collected through the data collection module, and the audit support data is preprocessed; wherein the audit support data includes basic data and flow data; Integrate the gantry transaction flow and the gantry brand recognition flow to generate a gantry flow model, and select the provincial boundary gantry flow model and the border gantry flow model based on the gantry flow model; In the process of integrating the gantry flow model, starting from the time interval, the data with time intervals within the threshold are merged and regarded as the same gantry passage, and the unmatched data are retained; in the process of screening the gantry flow model, by judging whether the path contains continuous outbound gantries and inbound gantries and the relative positions of the two, the scene of merging several cross-border passages in one passage is actually included; Integrate the lane entrance flow, provincial boundary gantry transaction flow, and border gantry transaction flow to generate an entrance flow model; The lane exit flow is associated with the entrance flow model, and the exit transactions are extracted from the provincial boundary gantry flow model and the gantry sign recognition flow, and then merged to generate the exit flow model; The exit flow model is used as a baseline for traversal, and the door frame license plate identifications queried are sorted and deduplicated according to time, and a license plate identification path charging unit combination is generated; The export flow model is used as a baseline for traversal, and after the queried gantry transactions are sorted and deduplicated according to time, a transaction path charging unit combination is generated, and a high-frequency path is obtained according to the transaction path charging unit combination; wherein the high-frequency path acquisition includes: Calculate the MD5 digest of the transaction path charging unit combination and form a key-value pair of MD5 digest-charging unit combination; form a "one-to-many" mapping relationship with "entry + exit" as the key and MD5 digest as the value; traverse the data of the previous step, perform aggregate statistics, and calculate the number of occurrences of each charging unit combination; traverse the data of the previous step again and calculate the high-frequency path; The similarity of the transaction path charging unit combination and the card recognition path charging unit combination is calculated and the audit fee amount is obtained; the audit suspicion is calculated based on the similarity threshold and the audit fee amount.

2. The high-speed audit system based on path algorithm according to claim 1 is characterized in that: By associating the gantry flow model with the gantry normative definition data, the provincial boundary gantry flow model and the border gantry flow model are screened out.

3. The high-speed audit system based on path algorithm according to claim 1 is characterized in that: The basic data include normative definition data of toll stations and gantries and normative definition data of toll units; the flow data include lane entrance flow, lane exit flow, gantry transaction flow and gantry plate recognition flow.

4. The high-speed audit system based on path algorithm according to claim 1 is characterized in that: The audit suspicion degree is obtained through an artificial intelligence model; wherein the artificial intelligence model includes a neural network model and a decision tree model.

5. The high-speed audit system based on path algorithm according to claim 1 is characterized in that: The vehicle information data in the flow data is extracted, and the vehicle information data is integrated and written into the vehicle model library; the vehicle model library counts the vehicle information data corresponding to each license plate that appears.

Images