A vehicle path reconstruction method based on license plate recognition and camera state

Abstract

The application discloses a vehicle path reconstruction method based on license plate recognition and a pipe state. The method can accurately restore the vehicle path and repair the traffic caused by missing detection or no pipe installation, and has good performance in actual cases. Compared with the prior art, the application has the following advantages: the data of the pipe can obtain a large number of key nodes passed by vehicles, and then the algorithm is used to connect the key nodes to obtain a path with high accuracy. In addition, the analysis based on the pipe data can directly obtain the OD of the motor vehicle, avoiding the error caused by the traditional division.

Description

Technical Field

[0001] This invention relates to a path reconstruction and checkpoint status recognition system based on license plate recognition data, belonging to the fields of pattern recognition and data mining technology. Background Technology

[0002] Reconstructing vehicle routes can reconstruct the travel demand characteristics of motor vehicles with high confidence, specifically obtaining traffic flow and travel routes within a region. Traditional methods are based on OD sampling, followed by determining motor vehicle travel demand using utility theory. However, the distribution of motor vehicle demand across the traffic network often requires accurate calibration of road characteristics and route selection parameters, a complex process that is difficult to significantly improve in terms of accuracy. The widespread deployment of checkpoints provides a new solution for understanding motor vehicle travel characteristics. However, considering the actual conditions of intersections and optimal placement strategies, there may be intersections without checkpoints in the entire road network. Furthermore, checkpoints may experience malfunctions or data loss during operation. Therefore, it is necessary to reconstruct continuous travel routes for motor vehicles based on existing checkpoint data. Current technologies typically select alternative routes by calculating the travel time of each route and matching it to the actual time difference. However, the travel time distribution of intermittent traffic flow under signal control does not satisfy the necessary and sufficient conditions for selecting alternative routes. The flow delay relationship of the road network differs during peak and off-peak hours, and current algorithms do not strictly distinguish between driving speeds and delays under different traffic conditions.

[0003] The above reasons may lead to discontinuous travel routes. Therefore, it is necessary to combine the traffic conditions on the road network to complete the splicing and reconstruction of routes and to restore traffic flow. Summary of the Invention

[0004] The technical problem this invention aims to solve is that existing methods for reconstructing traffic routes may lead to discontinuous travel paths.

[0005] To solve the above-mentioned technical problems, the technical solution of the present invention is to provide a vehicle path reconstruction method based on license plate recognition and checkpoint status, characterized by comprising the following steps:

[0006] Step 1: Obtain the topology of the road network through high-definition map acquisition, and at the same time obtain the license plate information of vehicles turning at each turn through checkpoint equipment, and then preprocess the collected data;

[0007] Step 2: Using checkpoint data, calculate the actual operating speed of neighboring nodes under various traffic conditions in different time periods. Based on the actual operating speed, estimate the travel time of neighboring nodes under various traffic conditions. The calculation of the actual operating speed of neighboring nodes under various traffic conditions using checkpoint data includes the following steps:

[0008] Step 2-1: Based on the road network topology and checkpoint locations, determine the set of road segments R_link_AVI where checkpoints are set at both upstream and downstream intersections;

[0009] Step 2-2: Based on the time difference, determine the set of travel times for neighboring nodes of each road segment in the road segment set R_link_AVI:

[0010]

[0011] In the formula, R Travle Time This represents the set of travel times of the neighboring nodes of a road segment in the road segment set R_link_AVI; This indicates the time when vehicle k passes through the downstream entrance lane; Indicates the time when vehicle k passes through the upstream entrance lane; t k This represents the travel time of vehicle k to its nearest neighbor nodes, in seconds.

[0012] Steps 2-3: Sort the travel times of neighboring nodes throughout the day from smallest to largest, assuming the top 10% correspond to vehicles operating in free-flow mode, and determine the free-flow speed for that road segment based on this. For example... Figure 1 As shown:

[0013]

[0014] In the formula: V0_link represents the free-flow state of the road segment; length_link represents the length of the road segment; t_min10% represents the top 10% of the travel time sets of the nearest road segments, arranged from smallest to largest.

[0015] Steps 2-4: Obtain the set R of travel times for neighboring nodes. Travle Time The quantile is 85%. Then, the minimum normal operating speed of each vehicle is calculated based on the length of the road segment of the adjacent node. Finally, the average is calculated to determine the operating speed between adjacent nodes.

[0016] Steps 2-5: Accurately divide time periods according to the traffic flow status during peak and off-peak hours, and give the actual speed distribution of each time period based on the running speed between adjacent nodes obtained in Step 2-4;

[0017] Step 3: Based on the same road or road level in the network, complete the travel times of neighboring nodes for road sections without checkpoints, and use the completed travel times of neighboring nodes as the initial weights of the updated road network;

[0018] Step 4: Determine the travel chain discrimination threshold based on the deployment density of checkpoints and the detected travel time of road segments. If a vehicle stops for an indefinite period of time in the road network and exceeds the set travel chain discrimination threshold, the travel chain of the current vehicle will be classified.

[0019] Step 5: Using the travel time of neighboring nodes as the generalized cost, use Dijkstra's algorithm to obtain the set of reconstructed paths;

[0020] Step 6: Determine the path reconstruction result under Dijkstra's algorithm, which includes the following steps:

[0021] Step 6-1: Based on the set reliability rate Robustness, use cluster analysis to determine the normal checkpoint set R_Normal;

[0022] Step 6-2: Determine the variance Dijkstra_V(FNR) and the average relative error Dijkstra_E(FNR) of the normal checkpoint set R_Normal under Dijkstra's algorithm;

[0023] Step 6-3: Record Dijkstra_V(FNR) and Dijkstra_E(FNR) to the reconstruction path determined by the minimum travel time, respectively;

[0024] Step 6-4: Calculate the global average confidence Dijkstra_E(FNR)_All and record it in the reconstruction path determined by the minimum travel time;

[0025] Step 7: Perform KSP to determine the K shortest paths, and combine the false negative rate to obtain the path with the maximum utility and confidence. This includes the following steps:

[0026] Step 7-1: According to KSP, select K=m, and propose m routes with the minimum travel time;

[0027] Step 7-2: Considering the high stability and deployment density of the checkpoints, the possibility of large-segment path loss is very low. Therefore, the false negative rate (FNR) is used to calculate the utility of each path, as shown in the following formula:

[0028]

[0029] in, This represents the product of the false negative rate (FNR) of the checkpoints along the path, where n represents the number of checkpoints along the path, and Travel_Time represents the travel time to neighboring nodes.

[0030] Step 7-3: Based on maximum utility U max The KSP algorithm is used to select a reasonable reconstruction path.

[0031] U max =max{U i}, i = 1, 2, ..., m

[0032] The KSP algorithm selects the maximum utility U.max The path is used as the reconstruction path;

[0033] Step 7-4: Save the paths obtained by the KSP algorithm, and use the mean of the false negative rate (FNR) as the confidence level of the path:

[0034] Confidence i =avg(FNR) i ), i = 1, 2, ..., n

[0035] Where n represents the number of neighboring nodes traversed on the current path.

[0036] Step 7-5: Based on the set reliability rate Robustness, use cluster analysis to determine the normal checkpoint set R_Normal;

[0037] Step 7-6: Determine the variance KSP_V(FNR) and mean relative error KSP_E(FNR) of the normal set under the KSP algorithm;

[0038] Step 7-7: Calculate the global average confidence KSP_E(FNR)_All and record the reconstruction path under the maximum utility determined by KSP;

[0039] Step 8: Based on the path determined by KSP, reverse the false negative rate and evaluate the working status of the checkpoint.

[0040] Preferably, in step 1, the collected data is preprocessed to supplement the detected data, thereby avoiding traffic loss and ensuring steering accuracy.

[0041] Preferably, step 3 includes the following steps:

[0042] Step 3-1: Determine the current link to be completed. i Determine the nearest link upstream or downstream where a checkpoint is located. j_ AVI;

[0043] Step 3-2: If link i and link j_ If AVI represents different segments of the same road, then the travel time difference is determined by the distance difference between the segments, as shown in the following formula:

[0044]

[0045] in, and Indicates road segment link i The travel time and length and Indicates road segment link j_AVI represents the travel time and length, and V0_ represents the free-flow velocity of the road segment.

[0046] Preferably, step 5 includes the following steps:

[0047] Step 5-1: Based on the time period divided by traffic conditions, retrieve the corresponding road network and weights;

[0048] Step 5-2: Calculate and save the shortest path according to Dijkstra's algorithm;

[0049] Step 5-3: Obtain the inlet flow rate through reconstruction: Dijkstra_ Reconstructed ;

[0050] Step 5-4: Obtain the relative error under Dijkstra's algorithm, i.e., the checkpoint false negative rate Dijkstra_, also denoted as FNR:

[0051]

[0052] IfDijkstra_ Reconstructed =0,thFNR=100%

[0053] In the formula, Q Detected This indicates the traffic data detected by the checkpoint;

[0054] Step 5-5: Generate the shortest travel time path according to Dijkstra's algorithm and calculate the false negative rate (FNR) at each checkpoint.

[0055] The method disclosed in this invention can accurately restore vehicle routes and traffic flow caused by missed detections or the absence of checkpoints. It has shown good performance in practical cases. Compared with existing technical solutions, this invention has the following specific advantages:

[0056] (1) This invention can obtain a large number of key nodes passed by vehicles through checkpoint data, and then use an algorithm to connect the key nodes to obtain a path with high accuracy. In addition, the analysis based on checkpoint data can directly obtain the OD of motor vehicles, avoiding the errors of traditional methods.

[0057] (2) This invention utilizes a two-step path reconstruction process. First, the working status of the checkpoint is initially determined, and then the path with the greatest utility is determined using KSP. Finally, based on the final path, the false negative rate of each checkpoint is calculated, which can be used to evaluate the reliability of the algorithm, reflect the poor data at the checkpoint, and evaluate the working status of each checkpoint by tracing the data back to its source.

[0058] (3) This invention is based on the assumption of maximum utility in user path selection, and then uses the working status of the checkpoint to infer the path selection. This part not only considers the utility of minimum cost, but also makes full use of the working status of the checkpoint to provide a high-confidence path selection.

[0059] (4) This invention converts the segment length of neighboring nodes in the traditional shortest path algorithm into travel time, which can better reflect the real-time generalized cost of the path (travel time). Based on this, it is recommended to accurately divide the data according to the fluctuation of traffic conditions, and then calculate the travel time of neighboring nodes in each time period. Attached Figure Description

[0060] Figure 1 The diagram illustrates the sample distribution of travel time for each road segment.

[0061] Figure 2 This diagram illustrates the travel time distribution of neighboring nodes (left: peak hours; right: off-peak hours).

[0062] Figure 3 This illustrates the calculation of initial weights based on travel time;

[0063] Figure 4 This illustrates the technical approach of the present invention;

[0064] Figure 5 This is a diagram illustrating the path selection process. Detailed Implementation

[0065] The present invention will be further illustrated below with reference to specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the teachings of this invention, those skilled in the art can make various alterations or modifications to the invention, and these equivalent forms also fall within the scope defined by the appended claims.

[0066] This embodiment discloses a vehicle route reconstruction method based on license plate recognition and checkpoint status, which includes the following steps:

[0067] Step 1: Obtain the road network topology using high-definition maps, and simultaneously acquire license plate information of vehicles turning at each direction using checkpoint equipment. Then, preprocess the collected data. In Step 1, supplementary data for unique detection is added to avoid data loss and ensure steering accuracy.

[0068] To improve the road network topology and data preprocessing, step 1 further includes the following steps:

[0069] Step 1-1: Set up nodes around the intranet to ensure that the node segments are reachable.

[0070] Steps 1-2: Repair road segments and node numbers, and define the accessibility of the road network.

[0071] Steps 1-3: Modify the topology and turning definitions.

[0072] Steps 1-4: Add attribute values ​​for road segments, such as road grade, length, free flow speed, etc.

[0073] Steps 1-5: No checkpoint equipment was installed at the newly added edge intersections. Considering the travel needs of external traffic, but the time for vehicles to return to the internal network is uncontrollable, it is necessary to study and consider forcibly interrupting external traffic flow and separating paths. Specifically, the determination of whether a road segment is associated with external conditions is based on whether the vehicle leaves the edge segment. In this embodiment, the definition of whether a road segment is associated with external conditions is shown in Table 1 below.

[0074] Table 1 External Circumstances Related to Road Sections

[0075] ID link_out Remark link 1 1 Drive towards the outer nodes link 2 0 Driving towards the internal node

[0076] Step 2: Calculate the actual operating speed of the adjacent nodes where the checkpoints are located under various traffic conditions by time period using checkpoint data.

[0077] In step 2, to avoid constructing paths with extremely low probability due to excessive matching of travel times in the original algorithm, this embodiment converts the segment lengths of neighboring nodes in the traditional shortest path algorithm into travel times. Furthermore, traffic conditions in the road network vary significantly across different time periods. Therefore, to better reflect the real-time generalized cost of paths, this embodiment first determines the time periods for traffic conditions and then calculates the travel times of detectable neighboring road segments based on vehicle data from checkpoint entrances. The specific calculation method for step 2 includes the following steps: Step 2-1: Based on the road network topology and checkpoint locations, determine the set of road segments R_link_AVI where checkpoints are set at both upstream and downstream intersections.

[0078] Step 2-2: Based on the time difference, determine the set of travel times for neighboring nodes of each road segment in the road segment set R_link_AVI:

[0079]

[0080] In the formula, R Travle Time This represents the set of travel times of the neighboring nodes of a road segment in the road segment set R_link_AVI; This indicates the time when vehicle k passes through the downstream entrance lane; Indicates the time when vehicle k passes through the upstream entrance lane; t k This represents the travel time of vehicle k to its nearest neighbor nodes, in seconds.

[0081] Steps 2-3: Sort the travel times of neighboring nodes throughout the day from smallest to largest, assuming the top 10% correspond to vehicles operating in free-flow mode, and determine the free-flow speed for that road segment based on this. For example... Figure 1 As shown:

[0082]

[0083] In the formula: V0_ represents the free-flow state of the road segment; length_ represents the length of the road segment; t_min10% represents the top 10% of the travel time sets of neighboring nodes arranged from smallest to largest.

[0084] Steps 2-4: Traffic control at intersections determines whether vehicles stop when passing through the intersection. The resulting intermittent traffic flow causes the travel time distribution of neighboring nodes to exhibit a bimodal characteristic, such as... Figure 1 As shown. Take the set R of travel times for neighboring nodes. Travle The quantile is 85%. Then, the minimum normal operating speed of each vehicle is calculated based on the length of the road segment of the adjacent node. Finally, the average is calculated to determine the operating speed between adjacent nodes.

[0085] Steps 2-5: Accurately divide time periods according to the traffic flow status during peak and off-peak hours, and calculate the travel time by combining the running speeds between neighboring nodes obtained in Step 2-4.

[0086] Step 3: Based on the same road or road class in the network, complete the travel time of neighboring nodes for road sections without checkpoints. This includes the following steps:

[0087] Step 3-1: Determine the current link to be completed. i Determine the nearest link upstream or downstream where a checkpoint is located. j_ AVI;

[0088] Step 3-2: If link i and link j_ If AVI represents different segments of the same road, then the travel time difference is determined by the distance difference between the segments, as shown in the following formula:

[0089]

[0090] in, and Indicates road segment Link i The travel time and length and Indicates road segment Link j_ AVI represents the travel time and length, and V0_ represents the free-flow velocity of the road segment.

[0091] Step 3-3: Calculate the travel time of neighboring nodes As the initial weights of the updated road network, such as Figure 2 As shown.

[0092] Step 4: Determine the travel chain identification threshold based on the checkpoint deployment density and detected road segment travel time. Vehicles that stop for varying durations within the network, exceeding the set travel chain time threshold, will be classified as part of a travel chain. The specific calculation method is as follows:

[0093] Step 4-1: Determine the travel chain segmentation threshold based on the travel time distribution of neighboring nodes. The travel time of neighboring nodes depends on the deployment density of checkpoints.

[0094] Step 4-2: Combine multi-purpose travel strategies to determine the vehicle's purpose and accurately divide the travel chain to avoid mistaking a parked vehicle for one that is still in motion in order to match its travel time.

[0095] Step 4-3: Set a trip chain segmentation threshold. Exceeding the set time threshold indicates that the vehicle has been parked for an extended period within the entire trip chain. This threshold can be determined based on the travel time of neighboring nodes. The trip chain segmentation threshold can be defined as the right boundary of the 95% confidence interval for the travel time of neighboring nodes.

[0096] Step 5: Using the travel time of neighboring nodes as the generalized cost, use Dijkstra's algorithm to obtain the set of reconstructed paths. This includes the following steps:

[0097] Step 5-1: Based on the time periods divided by traffic conditions, retrieve the corresponding networks and weights from Table 2 below:

[0098] Table 2. Fields related to the generalized cost of road sections.

[0099]

[0100] Step 5-2: Calculate the shortest path using Dijkstra's algorithm and save it to Table 3 below:

[0101] Table 3 Path table under the shortest path (cost) algorithm

[0102] Serial Number Fields Notes Remark 1 HPHM License plate number 2 START_TIME Start time 3 END_TIME End time 4 START_CROSS_NO starting point 5 END_CROSS_NO end 6 PATH_NODE Total number of intersections 7 LENGTH Total path length 8 TIME Total Time 9 REPAIR_NODE Total number of nodes repaired 10 PATH path 11 PATH_JGSJ Over time

[0103] Step 5-3: Obtain the inlet flow rate through reconstruction: Dijkstra_Q Reconstructed ;

[0104] Step 5-4: Relative Error (Checkpoint Miss Rate) FNR under Dijkstra's Algorithm:

[0105]

[0106] IfDijkstra_Reconstructed =0,thFNR=100%

[0107] In the formula, Q Detected This indicates the traffic data detected by the checkpoint.

[0108] Obviously, without checkpoints, Dijkstra = 100%.

[0109] Step 5-5: Generate the shortest (travel time) path using Dijkstra's algorithm and calculate the false negative rate (FNR) at each checkpoint. Combined with... Figure 3 Step 6: Determining the path reconstruction result under Dijkstra's algorithm, which specifically includes the following steps:

[0110] Step 6-1: Based on the set reliability rate Robustness, use cluster analysis to determine the normal checkpoint set R_Normal;

[0111] Step 6-2: Determine the variance Dijkstra_() and the mean relative error Dijkstra_() of the normal checkpoint set R_Normal under Dijkstra's algorithm;

[0112] Step 6-3: Record Dijkstra_() and Dijkstra_() in Iteration 1 as shown in Table 4 below;

[0113] Table 4 Global Statistics

[0114]

[0115] Step 6-4: Calculate the global average confidence level Dijkstra_(FNR)_ and record it in Iteration 1 shown in Table 4 above.

[0116] Step 7: Perform KSP to determine the K shortest paths, and combine the false negative rate to obtain the path with the maximum utility and confidence. This includes the following steps:

[0117] Step 7-1: According to KSP, select K=m, and propose m routes with the minimum travel time;

[0118] Step 7-2: Considering the high stability and deployment density of the checkpoints, the possibility of large-segment path loss is very low. Therefore, the false negative rate (FNR) is used to calculate the utility of each path, as shown in the following formula:

[0119]

[0120] in, This represents the product of the false negative rate (FNR) of the checkpoints along the path, where n represents the number of checkpoints along the path, and Travel_ represents the travel time to neighboring nodes.

[0121] Step 7-3: Based on maximum utility U max The KSP algorithm is used to select a reasonable reconstruction path.

[0122] U max =max{U i},=1,2,…m

[0123] The KSP algorithm selects the maximum utility U. max The path is used as the reconstruction path, such as Figure 4 As shown. Step 7-4: Save the path obtained by the KSP algorithm, and use the mean of the false negative rate (FNR) as the confidence level of the path:

[0124] Confidence i =vg(FNR) i ), i = 1, 2, ..., n

[0125] Where n represents the number of neighboring nodes traversed on the current path.

[0126] Step 7-5: Based on the set reliability rate Robustness, use cluster analysis to determine the normal checkpoint set R_Normal;

[0127] Step 7-6: Determine the variance KSP_V(FNR) and mean relative error KSP_E(FNR) of the normal set under the KSP algorithm;

[0128] Step 7-7: Calculate the global average confidence level KSP_()_ and record it in Iteration 2 of Table 4 above.

[0129] Step 8: Based on the path determined by KSP, reverse the false negative rate and evaluate the working status of the checkpoint.

[0130] The fact that a vehicle passes through multiple checkpoints but is not detected is an extremely rare event. The possible reasons for this situation are: (1) the equipment may be disconnected while the vehicle is in operation;

[0131] (2) The vehicle may engage in other behaviors, such as parking or leaving the intranet;

[0132] (3) Vehicles make U-turns, but the existing road network does not have U-turn facilities, causing vehicles to detour.

[0133] Objective and Positioning: Since the travel times for each route are likely to be very similar, all falling within a high confidence interval, travel time can be used for route determination and verification.

[0134] The technical advantages of the technical solutions provided in the above embodiments are as follows:

[0135] (1) The average travel time of intersections under the target time period is uniformly adopted as its generalized cost. Specifically, the travel time includes free-flow travel time, road segment delay time and average waiting time of intersection nodes. (2) The problems of Dijkstra's algorithm in calculating the travel time of unreachable road segments and paths are avoided by calculating and modifying the weights to avoid unreasonable situations such as excessively fast or slow speeds.

[0136] (3) Considering the travel time of the road segment, the irrationality of selecting the KSP optimal path based on the travel time was determined.

Claims

1. A vehicle path reconstruction method based on license plate recognition and camera status, characterized in that, The method comprises the following steps: Step 1: obtaining the topological structure of the road network through high-definition map collection, and obtaining the license plate information of each turning vehicle through the turret device, and then preprocessing the collected data; Step 2: calculating the real running speed of the adjacent nodes of the installed turret in various traffic states through the turret data, and calculating the travel time of the adjacent nodes of the installed turret in various traffic states based on the real running speed, wherein the calculation of the real running speed of the adjacent nodes of the installed turret in various traffic states through the turret data comprises the following steps: Step 2-1: determining the road section set R_link_AVI in which the upstream and downstream intersections are provided with turrets according to the topological structure of the road network and the layout position of the turrets; Step 2-2: determining the travel time distribution of the adjacent nodes of the vehicle passing through each road section in the road section set R_link_AVI according to the time difference, as shown in the following formula: In the formula, R TravleTime denotes the set of travel times of adjacent nodes of adjacent nodes of links in the set of links R_link_AVI; denotes the time at which vehicle k passes the downstream on-ramp; denotes the time at which vehicle k passes the upstream on-ramp; t k denotes the adjacent node travel time of vehicle k; Step 2-3: arranging the travel time set of the adjacent nodes in ascending order, and determining the free flow speed of the road section according to the fact that the vehicles corresponding to the first 10% of the set are in the free flow state: In the formula, V0_link represents the free flow state of the road section, length_link represents the length of the road section, and t_min10% represents the first 10% set arranged in ascending order of the travel time set of the adjacent nodes; Step 2-4: Obtain the travel time set R of the adjacent node TravkeTime The 85% quantile is then calculated, and the minimum value of the normal operating speed of each vehicle is calculated according to the length of the road segment of the adjacent node, and finally the average operating speed between adjacent nodes is determined. Step 2-5: accurately dividing the time period according to the traffic flow state of the peak and flat peak periods, and giving the real speed distribution of each period in combination with the running speed between the adjacent nodes obtained in step 2-4; Step 3: completing the travel time of the adjacent nodes of the road section on which no turret is installed according to the same road or road level in the network, and taking the completed travel time of the adjacent nodes as the updated initial weight of the road network; Step 4: determining the travel chain discrimination threshold according to the layout density of the turret and the detected road section travel time, and dividing the travel chain of the vehicle when the vehicle is parked for an indefinite length of time in the road network and exceeds the set travel chain time period discrimination threshold; Step 5: taking the travel time of the adjacent nodes as the generalized cost, and obtaining the reconstructed path set by using the Dijkstra algorithm; Step 6: path reconstruction result determination under the Dijkstra algorithm, specifically comprising the following steps: Step 6-1: determining the normal turret set R_Normal by using clustering analysis according to the set reliability Robustness; Step 6-2: determining the variance Dijkstra_V(FNR) and the average relative error Dijkstra_E(FNR) of the normal turret set R_Normal under the Dijkstra algorithm; Step 6-3: recording Dijkstra_V(FNR) and Dijkstra_E(FNR) to the reconstructed path determined according to the minimum travel time; Step 6-4: calculating the global average confidence Dijkstra_E(FNR)_All and recording it to the reconstructed path determined according to the minimum travel time; Step 7: Perform KSP, determine K shortest paths, combine the missed detection rate to get the path with the maximum utility and confidence, including the following steps: Step 7-1: According to KSP, select K=m, and draft m paths with the minimum travel time; Step 7-2: Considering the high stability and density of the layout, the possibility of missing a large section of the path is very low, so the missed detection rate FNR is used to calculate the utility of each path, as shown in the following formula: wherein, represents the product of the false negative rate FNR of the path passing through the intersection, n represents the number of intersections passed through by the path, and Travel_Time represents the travel time of the adjacent nodes. Step 7-3: According to the maximum utility U max , select a reasonable reconstruction path by using KSP algorithm: U max = max{U u}, i = 1, 2,... m The KSP algorithm selects the path with the maximum utility U as the reconstruction path. max The KSP algorithm selects the path with the maximum utility U as the reconstruction path. Step 7-4: Save the path obtained by KSP algorithm, and use the average of the missed detection rate FNR as the confidence of the path: Confidence i = avg(FNR i ), i = 1, 2, … n Where n represents the number of adjacent nodes passed under the current path; Step 7-5: According to the set reliability Robustness, determine the normal camera set R_Normal by cluster analysis; Step 7-6: Determine the variance KSP_V(FNR) and the average relative error KSP_E(FNR) of the normal set under KSP algorithm; Step 7-7: Calculate the global average confidence KSP_E(FNR)_All, and record the reconstructed path under the maximum utility determined by KSP; Step 8: According to the path determined by KSP, the missed detection rate is calculated again to evaluate the working state of the camera.

2. The vehicle path reconstruction method based on license plate recognition and lens state according to claim 1, wherein, In step 1, the collected data is preprocessed to supplement the detected data, avoiding traffic loss and ensuring the turning accuracy.

3. The vehicle path reconstruction method based on license plate recognition and lens state according to claim 1, characterized in that, The step 3 includes the following steps: Step 3-1: Determine the current to-be-completed intersection link i Determine the most proximate upstream or downstream set-socket link j_ AVI; Step 3-2: If link i and link j_ AVI belong to different segments of the same road, then the travel time difference is determined by the segment distance difference, as shown in the following equation: where and denote the travel time and length of a link Link i and denote the travel time and length of a link Link j_ AVI, and V0_link denotes the free-flow speed of the link.​ 4. The vehicle path reconstruction method based on license plate recognition and lens state according to claim 1, characterized in that, The step 5 includes the following steps: Step 5-1: According to the time period divided according to the traffic state, the corresponding road network and weight are retrieved; Step 5-2: Calculate and save the shortest path according to Dijkstra algorithm; Step 5-3: Get the detected flow of the import lane by reconfiguration: Dijkstra_Q Reconstructed ; Step 5-4: Obtain the relative error under Dijkstra algorithm, that is, the camera missed detection rate Dijkstra_FNR, which is also expressed as FNR: If Dijkstra_Q Reconstructed = 0, then FNR = 100% in the equation, Q Detected represents the flow data detected by the socket. Step 5-5: Generate the shortest travel time path according to Dijkstra algorithm, and calculate the missed detection rate FNR of each camera.

Images