A vehicle driving mileage estimation method based on graph signal data

Abstract

This invention discloses a vehicle mileage estimation method based on image and signaling data, relating to the field of vehicle mileage estimation. The method includes the following steps: loading vehicle image data and signaling data to generate vehicle trajectory data and signaling trajectory data respectively; calculating the long-term association relationship between the vehicle and the mobile terminal number using an image-signaling fusion calculation strategy; performing high-confidence determination and filtering the vehicle's corresponding vehicle network number by combining the vehicle's APN dictionary and internet traffic data; filtering and compressing the signaling trajectory data of the vehicle network number; calculating the initial vehicle mileage and correcting it using a dynamic correction algorithm to obtain the final vehicle mileage. This method integrates vehicle image and signaling data, accurately determines the long-term association relationship between the vehicle and the mobile terminal through an image-signaling fusion calculation model, and significantly improves the accuracy and efficiency of vehicle mileage estimation by utilizing high-confidence filtering, trajectory data filtering and compression, and a dynamic correction algorithm.

Description

Technical Field

[0001] This invention relates to the field of vehicle mileage estimation, and more specifically to a method for estimating vehicle mileage based on map data. Background Technology

[0002] Vehicle mileage estimation technology is a crucial component of intelligent transportation systems and big data analytics. By integrating multi-source data such as vehicle trajectories and mobile communication signaling, it provides key support for traffic management, vehicle monitoring, and insurance billing. With the widespread adoption of vehicle-to-everything (V2X) technology, high-precision and high-efficiency mileage estimation methods have become an urgent industry need.

[0003] Traditional methods often rely on a single data source, which has significant limitations. For example, while GPS data has high accuracy, its coverage is limited and it consumes a lot of energy; base station signaling data has wide coverage but low positioning accuracy, especially prone to drift errors in areas with sparse base stations. Existing technologies lack effective multi-source data fusion mechanisms, leading to inaccurate determination of the long-term relationship between vehicles and mobile terminals, directly affecting the reliability of subsequent mileage estimation.

[0004] Existing mileage estimation algorithms are mostly based on the linear accumulation of base station spacing, without fully considering base station positioning drift and nonlinear path deviation. For example, in dense urban areas or complex road scenarios, frequent base station handovers lead to trajectory point redundancy and positioning jumps. Existing static correction models cannot adaptively adjust errors, resulting in a significant increase in accumulated errors and a large deviation between the estimated mileage and the actual mileage traveled.

[0005] Unoptimized trajectory data compression methods (such as full trajectory point processing) result in high computational complexity and resource consumption. In addition, existing technologies employ simplistic filtering strategies for stationary trajectory points and ping-pong switching points (such as relying solely on time intervals), which cannot effectively reduce redundant data and fails to meet the low-latency requirements of large-scale real-time vehicle monitoring scenarios, thus limiting the practical application scope of the technology.

[0006] Therefore, how to design a vehicle mileage estimation method based on map and information data that can integrate multi-source data, dynamically correct positioning errors, and efficiently compress trajectories to improve the accuracy and stability of mileage estimation is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0007] In view of this, the present invention provides a vehicle mileage estimation method based on image and signaling data. It constructs a spatiotemporal correlation model by fusing vehicle image data and signaling data, and combines trajectory collision data with equal weighting operations to match the long-term association relationship between vehicles and terminal numbers. It adopts a dynamic correction algorithm to effectively compensate for base station positioning drift errors and improve mileage estimation accuracy. At the same time, based on spatiotemporal index compression and redundant trajectory point filtering strategies, it significantly reduces data computation complexity and achieves high coverage and low latency dynamic estimation of vehicle mileage.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] A method for estimating vehicle mileage based on map and information data includes the following steps:

[0010] S1. Load vehicle image data and signaling data, and generate vehicle trajectory data and signaling trajectory data respectively;

[0011] S2. Based on the vehicle trajectory data and signaling trajectory data, calculate the long-term relationship between the vehicle and the mobile terminal number using a graph-signal fusion calculation model;

[0012] S3. Make a high-confidence determination on the long-term association relationship, and filter out the vehicle network number corresponding to the vehicle by combining the vehicle APN dictionary and Internet traffic data.

[0013] S4. Filter and compress the signaling trajectory data of the vehicle network number to exclude stationary trajectory points and ping-pong switching trajectory points;

[0014] S5. Calculate the initial vehicle mileage based on the filtered and compressed signaling trajectory data, and correct the initial vehicle mileage using a dynamic correction algorithm to obtain the vehicle mileage.

[0015] Preferably, S1 includes:

[0016] Vehicle image data is concatenated in chronological order of acquisition to form vehicle trajectory data, which is a sequence of triples:

[0017] {<CAM_lon1,CAM_lat1,t1> ,..., <CAM_lon j ,CAM_lat j ,t j >,..., <CAM_lon n ,CAM_lat n ,t n >}, where CAM_lon j Represents the longitude of the j-th acquisition pole, CAM_lat j Represents the latitude of the j-th data collection pole, t jThis indicates the j-th data collection time, and n represents the total number of vehicle trajectory points;

[0018] Signaling data is concatenated according to base station attachment time to form signaling trajectory data, which is a triplet sequence:

[0019] {<BS_lon1,BS_lat1,t1> ,..., <BS_lon i ,BS_lat i ,t i >,..., <BS_lon m ,BS_lat m ,t m >}, where BS_lon i Represents the longitude of the i-th base station, BS_lat i Represents the latitude of the i-th base station, t i represents the attachment time of the i-th base station, and m represents the total number of signaling trajectory points.

[0020] Preferably, S2 includes:

[0021] S21. Construct a spatiotemporal index model based on vehicle trajectory data and signaling trajectory data, compress the trajectory data according to geospatial coding and time slices, and form a standardized spatiotemporal slice sequence;

[0022] S22. Filter the set of base stations near the vehicle pole, and generate trajectory collision data by combining the vehicle trajectory timestamp and the signaling trajectory timestamp;

[0023] S23. Perform equal weighting on mobile phone numbers from different operators, and combine trajectory collision data with the image-message fusion calculation model to calculate the equal weighted trajectory similarity, and output the association between mobile phone numbers and vehicles whose trajectory similarity exceeds the threshold.

[0024] S24. Iteratively optimize the model parameters to minimize the sum of the trajectory similarity rankings of vehicles and mobile terminal numbers, and output a high-confidence long-term association relationship.

[0025] Preferably, S3 includes:

[0026] S31. Calculate the difference sequence {Diff1, Diff2, …, Diff} between adjacent scores in the confidence score sequence. m S32. Calculate the mean μ of the difference sequence. diff and standard deviation σ diff Set the threshold cv=μ diff +σ diffS33. If a certain difference exceeds the threshold cv, the corresponding score is determined to be a significant cliff score, and vehicle code pairs with scores higher than the cliff score are selected as high-confidence vehicle code pairs; S34. Based on the vehicle APN dictionary, the APN field in the Internet traffic data is associated with the mobile terminal number of the high-confidence vehicle code pair to determine the vehicle network number corresponding to the vehicle.

[0027] Preferably, in step S34, the construction of the vehicle APN dictionary includes:

[0028] Collect vehicle-specific APN information provided by automakers, including APN name, operator code, and vehicle model;

[0029] The APN information is matched with the APN field of the internet traffic data to generate a mapping table between APN and vehicle model.

[0030] Preferably, S4 includes:

[0031] S41. Aggregate trajectory points that appear consecutively within the coverage area of ​​the same base station, and retain only the first and last trajectory points entering and leaving the coverage area of ​​the base station; S42. Calculate the average moving speed and the angle between the lines of three consecutive trajectory points. If the average moving speed is less than a preset speed threshold or the angle between the lines is less than the angle threshold, then filter out the intermediate trajectory points.

[0032] Preferably, in step S42, the formula for calculating the average moving speed is:

[0033]

[0034] Where d1 and d2 are the great circle distances between adjacent trajectory points, and Δt1 and Δt2 are the corresponding time intervals.

[0035] Preferably, in step S5, the dynamic correction algorithm is expressed as follows:

[0036]

[0037] Where D' represents the vehicle's mileage, D represents the initial vehicle mileage, m' represents the total number of trajectory points in the filtered and compressed signaling trajectory data, and α and β are scaling parameters.

[0038] Preferably, the formula for calculating the initial vehicle mileage D is:

[0039]

[0040] Among them, BS_X i BS_X i+1 BS_Y represents the longitude in radians of the i-th and (i+1)-th base stations, respectively. i BS_Y i+1Let represent the radians of the latitude of the i-th and (i+1)-th base stations, respectively, and R represent the Earth's radius.

[0041] Preferably, the scaling parameters α and β are in the range of (0,2) and are determined by fitting an error correction curve based on the error distribution between historical trajectory data and actual mileage; and optimizing the values ​​of α and β using the least squares method to minimize the corrected mileage error.

[0042] As can be seen from the above technical solution, compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0043] 1. This method integrates vehicle image data and signaling data, utilizes a spatiotemporal indexing model and trajectory collision data generation mechanism, and combines the geographical association between base stations and poles to effectively match the long-term companion relationship between vehicles and mobile terminals. Simultaneously, by incorporating a map-signal fusion calculation model and introducing equal-weighting operations to eliminate differences in operator networks, it ensures fairness in trajectory similarity calculation under different base station coverage conditions, significantly improving the accuracy of the association between vehicles and terminal numbers, and laying a reliable data foundation for subsequent mileage estimation.

[0044] 2. Based on the filtered and compressed signaling trajectory data, a dynamic correction formula is adopted, combining historical error distribution and least squares parameter optimization to adaptively adjust the initial mileage calculation results. This algorithm effectively compensates for base station positioning drift errors through the synergistic effect of geospatial distance calculation and nonlinear correction terms, making the corrected mileage closer to the actual driving distance and significantly improving estimation accuracy.

[0045] 3. The original trajectory data is compressed using a spatiotemporal slice indexing model, combined with stationary point aggregation and ping-pong switching point filtering, significantly reducing the number of redundant trajectory points. Simultaneously, based on the high coverage characteristics of real-time signaling data, it supports dynamic tracking of vehicle trajectories over a wide area, achieving low-latency and highly stable mileage estimation while ensuring computational efficiency, making it suitable for various application scenarios in urban roads and remote areas. Attached Figure Description

[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0047] Figure 1 A flowchart of a vehicle mileage estimation method based on image data is provided in an embodiment of the present invention;

[0048] Figure 2A schematic diagram illustrating the process of calculating the long-term relationship between vehicle and mobile terminal numbers provided in an embodiment of the present invention;

[0049] Figure 3 This is a schematic diagram illustrating the process of filtering out the vehicle network number corresponding to a vehicle, as provided in an embodiment of the present invention.

[0050] Figure 4 This is a schematic diagram illustrating the process of filtering and compressing signaling trajectory data of vehicle network numbers, as provided in an embodiment of the present invention. Detailed Implementation

[0051] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0052] like Figure 1 As shown, this embodiment provides a method for estimating vehicle mileage based on image data, including the following steps:

[0053] S1. Load vehicle image data and signaling data, and generate vehicle trajectory data and signaling trajectory data respectively;

[0054] S2. Based on the vehicle trajectory data and signaling trajectory data, calculate the long-term relationship between the vehicle and the mobile terminal number using a graph-signal fusion calculation model;

[0055] S3. Make a high-confidence determination on the long-term association relationship, and filter out the vehicle network number corresponding to the vehicle by combining the vehicle APN dictionary and Internet traffic data.

[0056] S4. Filter and compress the signaling trajectory data of the vehicle network number to exclude stationary trajectory points and ping-pong switching trajectory points;

[0057] S5. Calculate the initial vehicle mileage based on the filtered and compressed signaling trajectory data, and correct the initial vehicle mileage using a dynamic correction algorithm to obtain the vehicle mileage.

[0058] This method integrates vehicle imagery and signaling data to construct an image-signal fusion computing model, achieving high-precision long-term correlation matching between vehicles and terminal numbers. By combining high-confidence judgment and APN dictionary filtering methods, low-confidence interference data is filtered out, and redundant points are removed by compressing trajectory data, improving mileage estimation efficiency. Simultaneously, a dynamic correction algorithm is introduced to optimize parameters based on historical error distribution, significantly improving the accuracy and practicality of mileage estimation and providing reliable technical support for vehicle management.

[0059] The following provides a further detailed explanation of each step in the above implementation process;

[0060] In this embodiment S1, vehicle image data and signaling data are loaded to generate vehicle trajectory data and signaling trajectory data, respectively; specifically including:

[0061] Vehicle image data is concatenated in chronological order of acquisition to form vehicle trajectory data, which is a sequence of triples:

[0062] {<CAM_lon1,CAM_lat1,t1> ,..., <CAM_lon j ,CAM_lat j ,t j >,..., <CAM_lon n ,CAM_lat n ,t n >}, where CAM_lon j Represents the longitude of the j-th acquisition pole, CAM_lat j Represents the latitude of the j-th data collection pole, t j This indicates the j-th data collection time, and n represents the total number of vehicle trajectory points;

[0063] By concatenating triplet sequences in chronological order using vehicle image acquisition equipment, high-precision but limited-coverage vehicle trajectory data is generated. Table 1 shows the time series of vehicle IDs and pole positions, ensuring spatiotemporal consistency of the data;

[0064] Table 1

[0065] Vehicle ID Rod longitude pole latitude Timestamp v1 CAM_lon1 CAM_lat1 time1 v1 CAM_lon2 CAM_lat2 time2 ... ... ... ... v6 CAM_lon6 CAM_lat6 time6

[0066] Signaling data is concatenated according to base station attachment time to form signaling trajectory data, which is a triplet sequence:

[0067] {<BS_lon1,BS_lat1,t1> ,..., <BS_lon i ,BS_lat i ,t i >,..., <BS_lon m ,BS_lat m ,t m >}, where BS_lon i Represents the longitude of the i-th base station, BS_lat i Represents the latitude of the i-th base station, t i This represents the attachment time of the i-th base station, and m represents the total number of signaling trajectory points;

[0068] The triplet sequence generated based on base station attachment events has wide coverage but low accuracy; Table 2 below presents the handover records of mobile terminals at different base stations, providing a basis for subsequent correlation analysis;

[0069] Table 2

[0070] phone number Base station longitude Base station latitude Timestamp u1 BS_lon1 BS_lat1 time1 u1 BS_lon2 BS_lat2 time3 u1 BS_lon3 BS_lat3 time4 u1 BS_lon4 BS_lat4 time5 ... ... ... ... u6 BS_lon6 BS_lat6 time6

[0071] This step lays a standardized input foundation for subsequent fusion computing by structuring multi-source data, taking into account both high accuracy and wide coverage, and supporting accurate matching between vehicles and terminals.

[0072] In this embodiment S2, based on the vehicle trajectory data and signaling trajectory data, a long-term association relationship between the vehicle and the mobile terminal number is calculated using a graph-signal fusion calculation model; for example... Figure 2 As shown, it specifically includes:

[0073] S21. Construct a spatiotemporal index model based on vehicle trajectory data and signaling trajectory data, compress the trajectory data according to geospatial coding and time slices, and form a standardized spatiotemporal slice sequence;

[0074] S22. Filter the set of base stations near the vehicle pole, and generate trajectory collision data by combining the vehicle trajectory timestamp and the signaling trajectory timestamp; the purpose is to filter the spatiotemporally overlapping trajectory collision data by matching the vehicle pole timestamp and the base station attachment timestamp, and extract the spatiotemporal correlation features between the vehicle and the terminal.

[0075] S23. Perform equal weighting on mobile phone numbers from different operators, and combine trajectory collision data with the image-message fusion calculation model to calculate the equal weighted trajectory similarity, and output the association between mobile phone numbers and vehicles whose trajectory similarity exceeds the threshold.

[0076] Specifically, the equitable distribution operation in this step involves calculating the geographical distance between the vehicle pole and nearby operator base stations, grouping them by operator, and generating base station priority numbers based on distance. Then, the spatiotemporal collision data of the vehicle trajectory and the mobile terminal signaling trajectory are associated with this priority to eliminate the interference of differences in network coverage density among different operators on trajectory similarity calculation. Its core is to ensure the fairness and accuracy of vehicle and terminal trajectory matching in scenarios with uneven base station distribution or differences in operator coverage through distance sorting and group association mechanism, thereby improving the reliability of the association relationship determination.

[0077] Furthermore, in the image-message fusion computing model, the formula for calculating the similarity of equitable trajectories is as follows:

[0078]

[0079] Where TL = TR ∩ L, the intersection of trajectory and residence TL' = TR - L, |TL| and |TL'| represent the number of elements in TL and TL' respectively, the function bin represents the binning operation on the Rank in TL and TL', and w TL Weight of work and residence location, w TL’ Weighting of non-work-residence location, w i Ranking of work-residence locations by weighting, w j Weighting of non-work-residence location in the ranking;

[0080] The rationale behind this formula lies in its multi-scale assessment of trajectory similarity through spatial regional differential weighting and ranking binning: the trajectory points are divided into two categories, TL (workplace / residential) and TL' (non-workplace / residential), and each is assigned a weight w. TL and w TL’ This highlights the value of stable work-residence behavior patterns; simultaneously, it utilizes the binning function bin(⋅) combined with binning weights w i w j Discretizing the regional rankings takes into account both the spatial semantic characteristics of the trajectory and computational robustness. By adjusting the parameters, it can be flexibly adapted to different analysis scenarios, ultimately achieving accurate similarity measurement under the target.

[0081] S24. Iteratively optimize the model parameters to minimize the sum of the trajectory similarity rankings of vehicles and mobile terminal numbers, and output a high-confidence long-term association relationship.

[0082] Specifically, the obtained long-term associations include: the range of license plate numbers (vehicle IDs), mobile terminal numbers, and confidence scores, as shown in Table 3 below:

[0083] Table 3

[0084] Vehicle ID Mobile terminal number Confidence v1 u4 13 v1 u1 12 v1 u5 3 v1 u6 2 v1 u3 1 v1 u2 1

[0085] This step effectively reduces the complexity of data processing by constructing a spatiotemporal index model to compress trajectory data into a standardized spatiotemporal slice sequence. Subsequently, by filtering the set of base stations near the vehicle pole and combining them with timestamps to generate trajectory collision data, a preliminary spatiotemporal association between the vehicle and the mobile terminal number is achieved. The image-message fusion computing model also considers the differences in coverage of base stations from different operators, and eliminates the impact of these differences on trajectory similarity calculation through equalization operations. Finally, by iteratively optimizing the model parameters, a high-confidence long-term association relationship is output.

[0086] In this embodiment S3, a high-confidence determination is made on the long-term association relationship, and the vehicle network number corresponding to the vehicle is filtered out by combining the vehicle APN dictionary and internet traffic data; such as Figure 3 As shown, it specifically includes:

[0087] S31. Calculate the difference sequence {Diff1, Diff2, …, Diff} between adjacent scores in the confidence score sequence. m S32. Calculate the mean μ of the difference sequence. diff and standard deviation σ diff Set the threshold cv=μ diff +σ diff S33. If a certain difference exceeds the threshold cv, the corresponding score is determined to be a significant cliff score, and vehicle code pairs with scores higher than the cliff score are selected as high-confidence vehicle code pairs; S34. Based on the vehicle APN dictionary, the APN field in the Internet traffic data is associated with the mobile terminal number of the high-confidence vehicle code pair to determine the vehicle network number corresponding to the vehicle.

[0088] The construction of the vehicle APN dictionary includes:

[0089] Collect vehicle-specific APN information provided by automakers, including APN name, operator code, and vehicle model;

[0090] The APN information is matched with the APN field of the internet traffic data to generate a mapping table between APN and vehicle model.

[0091] In this step, the mean difference is 2.4 and the standard deviation is approximately 3.7. Therefore, only u1 and u4 meet the high confidence condition. Further combining internet traffic data and vehicle APN dictionary, the vehicle network number corresponding to vehicle v1 is selected as u1. Through statistical analysis and APN mapping, low-confidence interference data is effectively filtered out, the vehicle network number corresponding to the vehicle is locked, and the reliability of subsequent mileage estimation is improved.

[0092] In this embodiment S4, the signaling trajectory data of the vehicle network number is filtered and compressed to exclude stationary trajectory points and ping-pong switching trajectory points; for example Figure 4 As shown, it specifically includes:

[0093] S41. Aggregate trajectory points that appear consecutively within the coverage area of ​​the same base station, retaining only the first and last trajectory points entering and leaving the coverage area of ​​the base station; S42. Calculate the average moving speed and the angle between the lines connecting three consecutive trajectory points. If the average moving speed is less than a preset speed threshold or the angle between the lines is less than an angle threshold, then filter out the intermediate trajectory points; Specifically, the preset speed threshold ranges from 4 to 8 km / h, and the angle threshold ranges from 10° to 40°.

[0094] Furthermore, the formula for calculating average moving speed is:

[0095]

[0096] Where d1 and d2 are the great circle distances between adjacent trajectory points, and Δt1 and Δt2 are the corresponding time intervals.

[0097] Specifically, the filtered and compressed signaling trajectory data is shown in Table 4 below:

[0098] Table 4

[0099] phone number Base station longitude Base station latitude Timestamp u1 BS_lon1 BS_lat1 time1 u1 BS_lon2 BS_lat2 time3 u1 BS_lon4 BS_lat4 time5

[0100] This step mainly filters and compresses the signaling trajectory data of the vehicle network number to exclude stationary trajectory points and ping-pong handover trajectory points, thereby improving the accuracy and efficiency of mileage estimation. By aggregating trajectory points that appear continuously within the coverage area of ​​the same base station and calculating the average moving speed and the angle between the connecting lines of the continuous trajectory points, redundant and abnormal trajectory points can be effectively filtered out, while retaining key information.

[0101] In this embodiment S5, the initial vehicle mileage is calculated based on the filtered and compressed signaling trajectory data, and the initial vehicle mileage is corrected by a dynamic correction algorithm to obtain the vehicle mileage.

[0102] The dynamic correction algorithm is expressed as follows:

[0103]

[0104] Where D' represents the vehicle's mileage, D represents the initial vehicle mileage, m' represents the total number of trajectory points in the filtered and compressed signaling trajectory data, and α and β are scaling parameters.

[0105] Furthermore, the formula for calculating the initial vehicle mileage D is as follows:

[0106]

[0107] Among them, BS_X i BS_X i+1 BS_Y represents the longitude in radians of the i-th and (i+1)-th base stations, respectively. i BS_Y i+1 Let represent the radians of the latitude of the i-th and (i+1)-th base stations, respectively, and R represent the Earth's radius.

[0108] In addition, the scaling parameters α and β are in the range of (0,2) and are determined by fitting an error correction curve based on the error distribution between historical trajectory data and actual mileage; and optimizing the values ​​of α and β by least squares method to minimize the corrected mileage error.

[0109] Here, based on specific vehicle trajectory data and signaling trajectory data, this step will be further explained;

[0110] If BS_lon1=111.83, BS_lon2=111.9, BS_lon4=111.95, BS_lat1=29.22, BS_lat2=29.22, BS_lat4=29.24, and the measured actual driving distance is 19.5 kilometers, then the steps for estimating the driving mileage based on the dynamic correction algorithm are as follows:

[0111] 1) Calculate the radians: BS_X1=1.9518, BS_Y1=0.51, BS_X2=1.953, BS_Y2=0.51, BS_X4=1.9539, BS_Y4=0.5103

[0112] 2) D1 = 2 × 6371 ×

[0113]

[0114] D2 = 2 × 6371 ×

[0115]

[0116] The initial total distance D = D1 + D2 = 6.79 + 5.34 = 12.13 kilometers.

[0117] 3) Let α = 1, β = 0.5,

[0118] The corrected total mileage D' = 12.13 × (1 + tan(12.13 / 3) × 0.5) = 19.8. It can be seen that compared with the initial total mileage D (12.13 km) calculated directly using the base station spacing, the corrected total mileage D' (19.8 km) is closer to the actual driving route of 19.5 km.

[0119] In this example, the corrected mileage error was only 1.5%, which verifies the comprehensive advantages of the method in terms of accuracy, real-time performance and coverage, and provides reliable technical support for intelligent traffic management and vehicle networking services.

[0120] This embodiment constructs an image-signal fusion computing model by fusing vehicle imagery and signaling data, achieving high-precision long-term association matching between vehicles and terminal numbers. Combining high-confidence judgment and APN dictionary filtering methods, it successfully filters out the vehicle network numbers corresponding to the vehicles and effectively filters and compresses the signaling trajectory data. Finally, the initial vehicle mileage is corrected through a dynamic correction algorithm, significantly improving the accuracy and practicality of mileage estimation and providing reliable technical support for vehicle management.

[0121] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the systems disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the descriptions are relatively simple; relevant parts can be referred to the method section.

[0122] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for estimating vehicle mileage based on map and information data, characterized in that, Includes the following steps: S1. Load vehicle image data and signaling data, and generate vehicle trajectory data and signaling trajectory data respectively; S2. Based on the vehicle trajectory data and signaling trajectory data, calculate the long-term relationship between the vehicle and the mobile terminal number using a graph-signal fusion calculation model; include: S21. Construct a spatiotemporal index model based on vehicle trajectory data and signaling trajectory data. Compress the trajectory data according to geospatial coding and time slices to form a standardized spatiotemporal slice sequence. S22. Filter the set of base stations near the vehicle pole and generate trajectory collision data by combining the vehicle trajectory timestamp and the signaling trajectory timestamp. S23. Perform equal weighting operations on mobile phone numbers from different operators and input the trajectory collision data into the image-signal fusion calculation model to calculate the equal weighted trajectory similarity. Output the association between mobile phone numbers with trajectory similarity exceeding a threshold and the vehicle. S24. Iteratively optimize the model parameters to minimize the sum of the trajectory similarity rankings of the vehicle and the mobile terminal number, and output a high-confidence long-term association relationship. S3. Make a high-confidence determination on the long-term association relationship, and filter out the vehicle network number corresponding to the vehicle by combining the vehicle APN dictionary and Internet traffic data. S4. Filter and compress the signaling trajectory data of the vehicle network number to exclude stationary trajectory points and ping-pong switching trajectory points; S5. Calculate the initial vehicle mileage based on the filtered and compressed signaling trajectory data, and correct the initial vehicle mileage using a dynamic correction algorithm to obtain the vehicle mileage.

2. The method for estimating vehicle mileage based on image data according to claim 1, characterized in that, S1 includes: Vehicle image data is concatenated in chronological order of acquisition to form vehicle trajectory data, which is a sequence of triples: {<CAM_lon1,CAM_lat1,t1> ,..., <CAM_lon j ,CAM_lat j ,t j >,..., <CAM_lon n ,CAM_lat n ,t n >}, where CAM_lon j Represents the longitude of the j-th acquisition pole, CAM_lat j Represents the latitude of the j-th data collection pole, t j This indicates the j-th data collection time, and n represents the total number of vehicle trajectory points; Signaling data is concatenated according to base station attachment time to form signaling trajectory data, which is a triplet sequence: {<BS_lon1,BS_lat1,t1> ,..., <BS_lon i ,BS_lat i ,t i >,..., <BS_lon m ,BS_lat m ,t m >}, where BS_lon i Represents the longitude of the i-th base station, BS_lat i Represents the latitude of the i-th base station, t i represents the attachment time of the i-th base station, and m represents the total number of signaling trajectory points.

3. The method for estimating vehicle mileage based on image data according to claim 1, characterized in that, S3 includes: S31. Calculate the difference sequence {Diff1, Diff2, ..., Diff...} between adjacent scores in the confidence score sequence. m }; S32. Calculate the mean μ of the difference sequence. diff and standard deviation σ diff Set the threshold cv=μ diff +σ diff , where k=2; S33. If a certain difference exceeds the threshold cv, the corresponding score is determined to be a significant cliff score, and vehicle code pairs with scores higher than the cliff score are selected as high-confidence vehicle code pairs. S34. Based on the vehicle APN dictionary, associate the APN field in the internet traffic data with the mobile terminal number of the high-confidence vehicle code pair to determine the vehicle network number corresponding to the vehicle.

4. The method for estimating vehicle mileage based on image data according to claim 3, characterized in that, In step S34, the construction of the vehicle APN dictionary includes: Collect vehicle-specific APN information provided by automakers, including APN name, operator code, and vehicle model; The APN information is matched with the APN field of the internet traffic data to generate a mapping table between APN and vehicle model.

5. The method for estimating vehicle mileage based on image data according to claim 1, characterized in that, S4 includes: S41. Aggregate trajectory points that appear consecutively within the coverage area of ​​the same base station, and retain only the first and last trajectory points entering and leaving the coverage area of ​​the base station. S42. Calculate the average moving speed and the angle between the lines connecting three consecutive trajectory points. If the average moving speed is less than a preset speed threshold or the angle between the lines is less than an angle threshold, then filter out the intermediate trajectory points.

6. The method for estimating vehicle mileage based on image data according to claim 5, characterized in that, In step S42, the formula for calculating the average moving speed is: ； Where d1 and d2 are the great circle distances between adjacent trajectory points, and Δt1 and Δt2 are the corresponding time intervals.

7. The method for estimating vehicle mileage based on image data according to claim 1, characterized in that, In S5, the dynamic correction algorithm is expressed as follows: ； Where D' represents the vehicle's mileage, D represents the initial vehicle mileage, m' represents the total number of trajectory points in the filtered and compressed signaling trajectory data, and α and β are scaling parameters.

8. The method for estimating vehicle mileage based on image data according to claim 7, characterized in that, The formula for calculating the initial vehicle mileage D is as follows: ； Among them, BS_X i BS_X i+1 BS_Y represents the longitude in radians of the i-th and (i+1)-th base stations, respectively. i BS_Y i+1 Let represent the radians of the latitude of the i-th and (i+1)-th base stations, respectively, and R represent the Earth's radius.

9. The method for estimating vehicle mileage based on image data according to claim 7, characterized in that, The scaling parameters α and β range from (0, 2) and are determined in the following way: Based on the error distribution between historical trajectory data and actual mileage, an error correction curve is fitted. The values ​​of α and β are optimized using the least squares method to minimize the corrected mileage error.

Images