A target positioning correction method and system applicable to vehicle-mounted inspection

Abstract

This invention provides a target positioning correction method and system applicable to vehicle-mounted inspections. The method includes: acquiring the installation error angle of a GPS device; calculating the north direction angle using the installation error angle; a camera on the inspection vehicle tracking multiple inspection targets separately using a target tracking algorithm to obtain multiple sets of raw GPS data for each target; calculating multiple sets of target intersection coordinates for each target using the multiple sets of raw GPS data, the north direction angle, and the GPS trajectory data of the inspection vehicle; and clustering the multiple sets of target intersection coordinates for each target to calculate the center point position of each target. This invention considers multiple factors affecting positioning accuracy and achieves accurate positioning of the inspection targets through installation correction, target tracking, and clustering algorithms, effectively improving target positioning accuracy to 0-3m.

Description

Technical Field

[0001] This invention relates to the field of target localization algorithms, and in particular to a target localization correction method and system applicable to vehicle-mounted inspection. Background Technology

[0002] Urban inspection methods include manual inspection and vehicle-mounted intelligent inspection. Manual inspection primarily involves personnel driving around to inspect facilities and events, recording and photographing information using qualification forms or mobile devices, and then compiling and reporting the data. This method is inefficient, costly, and lacks real-time performance. Vehicle-mounted intelligent inspection, on the other hand, deploys cameras and other sensors on inspection vehicles to capture real-time images of the surrounding environment while the vehicle is in motion. This data is then transmitted to a backend system for identification via mobile storage devices or wireless networks. With the development of technologies such as the Internet of Things (IoT), artificial intelligence (AI), and sensors, vehicle-mounted intelligent inspection systems, based on lightweight vehicle-mounted equipment and AI algorithms, are increasingly being applied to areas such as urban component data collection, road health inspection, and infrastructure inspection. These systems offer advantages such as high efficiency, low labor costs, and strong real-time performance, resulting in significant social and economic benefits.

[0003] Existing vehicle-mounted binocular intelligent inspection systems collect various types of raw data during inspections, such as vehicle trajectory, heading, speed, images and videos captured by the binocular cameras, and target recognition types, depth of field, angles, and recognition areas. Due to equipment installation and system limitations, these data are subject to errors, such as GPS and binocular camera installation errors, and target positioning calculation errors, which severely affect the positioning accuracy of the inspected targets.

[0004] The above content is only used to help understand the technical solution of the present invention and does not represent an admission that the above content is prior art. Summary of the Invention

[0005] To solve the above-mentioned technical problems, the present invention provides a target positioning correction method applicable to vehicle-mounted inspection, comprising:

[0006] S1: The inspection vehicle conducts inspections according to the preset route, obtains the GPS installation error angle, and calculates the north direction angle using the installation error angle;

[0007] S2: The camera on the inspection vehicle tracks multiple inspection targets separately using a target tracking algorithm, and obtains multiple sets of raw GPS data for each inspection target;

[0008] S3: Calculate the coordinates of multiple target intersection points for each inspection target by using multiple sets of raw GPS data, north direction angle, and GPS trajectory data of the inspection vehicle for each inspection target.

[0009] S4: Cluster the coordinates of multiple sets of intersection points of each inspection target to calculate the center point position of each inspection target.

[0010] Preferably, step S1 specifically includes:

[0011] S11: Obtain the angles A1, A2, ..., A between multiple trajectory points of the inspection vehicle. m and GPS heading angles B1, B2, ..., B m ;

[0012] S12: Calculate the GPS installation error angle using the following formula:

[0013] C r =(A r –B r +720) / 360

[0014] C av =(C 1+ C2+...+C m ) / m

[0015] Where m represents the m sets of data collected, C r C represents the interpolation result for each group. av The installation error angle for GPS;

[0016] S13: After installation error angle correction, the angle between the inspection target direction and the due north direction is obtained by calibration, and the north direction angle is obtained.

[0017] Preferably, step S2 specifically includes:

[0018] S21: The camera on the inspection vehicle captures an image at a preset frequency;

[0019] S22: Obtain the recognition area of ​​each inspection target in the image, and the angle A between the center point of the recognition area S and the straight line formed by the camera and the driving direction;

[0020] S23: In a continuous acquisition frame, if A i >A i-1 And S i >S i-1 If multiple targets appear simultaneously, they are identified as the same target, where i is the number of the acquisition frame. When multiple targets appear simultaneously, they are distinguished according to the distance depth L, where the distance depth L is the distance between the target image pixel and the camera.

[0021] S24: Track each inspection target separately and obtain multiple sets of raw GPS data for each inspection target.

[0022] Preferably, step S3 specifically includes:

[0023] S31: Extract two adjacent trajectory points of the inspection vehicle from the GPS trajectory data of the inspection vehicle. The coordinates of trajectory point Point(j) are (X... j ,Y j The direction of travel is R. j The angle between the binoculars and the target q being inspected is W. j The coordinates of the trajectory point Point(j+1) are (X... j+1 ,Y j+1 The direction of travel is R. j+1 The angle between the binoculars and the target q being inspected is W. j+1 j is the number of the inspection vehicle's trajectory point, and q is the number of the inspection target;

[0024] S32: Calculate the first angle angle1 and the second angle angle2 of the inspection target q. The calculation formula is as follows:

[0025] angle1=(R j +W j +720) / 360

[0026] angle2=(R j+1 +W j+1 +720) / 360;

[0027] S33: Calculate the coordinates of the j-th intersection point of the inspected target q using the first angle angle1 and the second angle angle2. j V j The calculation formula is:

[0028]

[0029]

[0030] S33: Repeat S31-S33 to obtain the coordinates of all target intersection points of the inspection target q.

[0031] Preferably, step S4 specifically includes:

[0032] S41: Divide the target intersection coordinates of the inspection target q into k groups, and randomly select p target intersection coordinates as cluster centers, where the value of p is less than k.

[0033] S42: Calculate the distance between the coordinates of each target intersection point and each cluster center, and assign the coordinates of each target intersection point to the nearest cluster center;

[0034] S43: The cluster centers and the coordinates of the assigned target intersection points constitute a cluster;

[0035] S44: For a cluster, each time a target intersection point coordinate is assigned, the cluster center will be reset according to the target intersection point coordinates in the current cluster.

[0036] S45: Repeat S41-S44 until the cluster centers in each cluster no longer change;

[0037] S46: Set weights based on the effectiveness of depth values. The total number of target intersection coordinates for the inspected target q is N, and the maximum depth value is Dep. max For the target intersection point coordinates (U) j V j The weight of the coordinates of the j-th target intersection point is:

[0038] T j =N*Dep max / Dep j

[0039] Among them Dep j Let N be the depth value of the coordinates of the j-th target intersection point; the maximum value of j is N.

[0040] S47: Calculate the total weight of the inspection target q: T = T1 + T2 + ... + T N ;

[0041] S48: Calculate the center point position (U,V) of the inspection target q. The calculation formula is as follows:

[0042] U = (U1*T1 + U2*T1 + ... + U N *T N ) / T;

[0043] V = (V1*T1 + V2*T1 + ... + V) N *T N ) / T.

[0044] A target positioning correction system suitable for vehicle-mounted inspection includes:

[0045] The north direction angle acquisition module is used by the inspection vehicle to obtain the GPS installation error angle when it conducts inspections according to a preset route, and to calculate the north direction angle based on the installation error angle.

[0046] The target tracking module is used by the camera on the inspection vehicle to track multiple inspection targets separately using a target tracking algorithm, and obtain multiple sets of raw GPS data for each inspection target;

[0047] The target intersection coordinate acquisition module is used to calculate and obtain multiple sets of target intersection coordinates for each inspection target by using multiple sets of original GPS data, north direction angle and GPS trajectory data of each inspection target;

[0048] The center point location acquisition module is used to cluster the coordinates of multiple sets of intersection points of each inspection target and calculate the center point location of each inspection target.

[0049] The present invention has the following beneficial effects:

[0050] Considering various factors affecting positioning accuracy, the system uses installation calibration, target tracking, and clustering algorithms to correct each item and achieve precise positioning of the inspection target, improving the target positioning accuracy to 0-3m. This enhances the practicality and real-time performance of the vehicle-mounted intelligent inspection system and improves the accuracy of accident positioning and processing efficiency. Attached Figure Description

[0051] Figure 1 This is a flowchart of a method according to an embodiment of the present invention;

[0052] Figure 2 This is a schematic diagram of the camera tracking the inspection target;

[0053] Figure 3 This is a schematic diagram for obtaining the coordinates of the target intersection point;

[0054] Figure 4 A schematic diagram of the clustering results for the target intersection point coordinates;

[0055] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0056] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0057] Reference Figure 1 This invention provides a target positioning correction method applicable to vehicle-mounted inspection, comprising:

[0058] S1: The inspection vehicle conducts inspections according to the preset route, obtains the GPS installation error angle, and calculates the north direction angle using the installation error angle;

[0059] S2: The camera on the inspection vehicle tracks multiple inspection targets separately using a target tracking algorithm, and obtains multiple sets of raw GPS data for each inspection target;

[0060] S3: Calculate the coordinates of multiple target intersection points for each inspection target by using multiple sets of raw GPS data, north direction angle, and GPS trajectory data of the inspection vehicle for each inspection target.

[0061] S4: Cluster the coordinates of multiple sets of intersection points of each inspection target to calculate the center point position of each inspection target.

[0062] In this embodiment, step S1 specifically includes:

[0063] S11: Obtain the angles A1, A2, ..., A between multiple trajectory points of the inspection vehicle. m and GPS heading angles B1, B2, ..., B m ;

[0064] Specifically, the angles between multiple trajectory points and the GPS heading angle are collected when the inspection vehicle is running stably at a speed greater than 20 km / h to ensure the accuracy of the calculation results.

[0065] S12: Calculate the GPS installation error angle using the following formula:

[0066] C r =(A r –B r +720) / 360

[0067] C av =(C 1+ C2+...+C m ) / m

[0068] Where m represents the m sets of data collected, C r C represents the interpolation result for each group. av C is the installation error angle for GPS. av The average interpolation result is the installation error angle between the GPS dual antennas and the driving direction of the inspection vehicle;

[0069] S13: After installation error angle correction, the angle between the inspection target direction and the due north direction is obtained by calibration, and the north direction angle is obtained.

[0070] In this embodiment, when tracking multiple inspection targets, it is necessary to set the camera's data acquisition frequency, collect inspection target data during vehicle movement, obtain all inspection targets within the visual range, and mark them; the collected inspection target data includes all identification data of all inspection targets, and the number of times the same target is identified is equal to the product of the system's acquisition frequency and acquisition time; targets with the same mark are judged based on the angle, recognition area, and center point trajectory change pattern; from the moment the inspection target with the same mark enters the frame to its final departure from the frame, the same inspection target is tracked, and multiple sets of raw GPS data of the same marked inspection target object from multiple observation angles are obtained; such as Figure 2 The image shows a schematic diagram of a vehicle tracking an inspection target while it is in motion.

[0071] Step S2 is as follows:

[0072] S21: The camera on the inspection vehicle captures an image at a preset frequency;

[0073] S22: Obtain the recognition area of ​​each inspection target in the image, and the angle A between the center point of the recognition area S and the straight line formed by the camera and the driving direction;

[0074] S23: In a continuous acquisition frame, if A i >A i-1 And S i >S i-1 If multiple targets appear simultaneously, they are identified as the same target, where i is the number of the acquisition frame. When multiple targets appear simultaneously, they are distinguished according to the distance depth L, where the distance depth L is the distance between the target image pixel and the camera.

[0075] S24: Track each inspection target separately and obtain multiple sets of raw GPS data for each inspection target.

[0076] In this embodiment, the original GPS data of the inspection target includes: the coordinates of the inspection vehicle at a certain trajectory point, the driving direction at that time, and the angle between the binoculars identifying the inspection target. Based on the acquired point and angle data of multiple identifications of the same inspection target, the intersection point between two non-parallel rays at different angles is calculated to obtain multiple sets of target intersection point coordinates for the inspection target object. For the case where an inspection target is only identified once, data calibration is performed by manually inputting parameters based on constant relationships; such as Figure 3 This is a schematic diagram illustrating how to obtain the coordinates of the first target intersection point using trajectory points Point1 and Point2. Figure 3 The intersection point is the coordinate of the first target intersection point;

[0077] Step S3 is as follows:

[0078] S31: Extract two adjacent trajectory points of the inspection vehicle from the GPS trajectory data of the inspection vehicle. The coordinates of trajectory point Point(j) are (X... j ,Y j The direction of travel is R. j The angle between the binoculars and the target q being inspected is W. j The coordinates of the trajectory point Point(j+1) are (X... j+1 ,Y j+1 The direction of travel is R. j+1 The angle between the binoculars and the target q being inspected is W. j+1 j is the number of the inspection vehicle's trajectory point, and q is the number of the inspection target;

[0079] S32: Calculate the first angle angle1 and the second angle angle2 of the inspection target q. The calculation formula is as follows:

[0080] angle1=(R j +W j+720) / 360

[0081] angle2=(R j+1 +W j+1 +720) / 360;

[0082] S33: Calculate the coordinates of the j-th intersection point of the inspected target q using the first angle angle1 and the second angle angle2. j V j The calculation formula is:

[0083]

[0084]

[0085] S33: Repeat S31-S33 to obtain the coordinates of all target intersection points of the inspection target q.

[0086] In this embodiment, the clustering in step S4 uses the open-source kmenes library. Step S4 specifically involves:

[0087] S41: Divide the target intersection coordinates of the inspection target q into k groups, and randomly select p target intersection coordinates as cluster centers, where the value of p is less than k.

[0088] S42: Calculate the distance between the coordinates of each target intersection point and each cluster center, and assign the coordinates of each target intersection point to the nearest cluster center;

[0089] S43: The cluster centers and the coordinates of the assigned target intersection points constitute a cluster;

[0090] S44: For a cluster, each time a target intersection point coordinate is assigned, the cluster center will be reset according to the target intersection point coordinates in the current cluster.

[0091] S45: Repeat S41-S44 until the cluster centers in each cluster no longer change; after clustering is complete, remove outliers. The clustering results are as follows: Figure 4 As shown;

[0092] S46: Set weights based on the effectiveness of depth values. The total number of target intersection coordinates for the inspected target q is N, and the maximum depth value is Dep. max For the target intersection point coordinates (U) j V j The weight of the coordinates of the j-th target intersection point is:

[0093] T j =N*Dep max / Dep j

[0094] Among them Depj Let N be the depth value of the coordinates of the j-th target intersection point; the maximum value of j is N.

[0095] S47: Calculate the total weight of the inspection target q: T = T1 + T2 + ... + T N ;

[0096] S48: Calculate the center point position (U,V) of the inspection target q. The calculation formula is as follows:

[0097] U = (U1*T1 + U2*T1 + ... + U N *T N ) / T;

[0098] V = (V1*T1 + V2*T1 + ... + V) N *T N ) / T.

[0099] This invention provides a target positioning correction system suitable for vehicle-mounted inspection, comprising:

[0100] The north direction angle acquisition module is used by the inspection vehicle to obtain the GPS installation error angle when it conducts inspections according to a preset route, and to calculate the north direction angle based on the installation error angle.

[0101] The target tracking module is used by the camera on the inspection vehicle to track multiple inspection targets separately using a target tracking algorithm, and obtain multiple sets of raw GPS data for each inspection target;

[0102] The target intersection coordinate acquisition module is used to calculate and obtain multiple sets of target intersection coordinates for each inspection target by using multiple sets of original GPS data, north direction angle and GPS trajectory data of each inspection target;

[0103] The center point location acquisition module is used to cluster the coordinates of multiple sets of intersection points of each inspection target and calculate the center point location of each inspection target.

[0104] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0105] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments. In the unit claims listing several devices, several of these devices may be embodied by the same hardware item. The use of the terms first, second, and third, etc., does not indicate any order and can be interpreted as identifiers.

[0106] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A target positioning correction method for vehicle-mounted inspection, characterized in that, The method comprises the following steps: S1: the inspection vehicle inspects according to a preset route, obtains an installation error angle of a GPS, and calculates a north direction angle through the installation error angle; S2: a camera on the inspection vehicle respectively tracks a plurality of inspection targets through a target tracking algorithm, and obtains a plurality of groups of original GPS data of the inspection targets; S3: a plurality of groups of target intersection coordinates of each inspection target are calculated through the plurality of groups of original GPS data of each inspection target, the north direction angle and GPS track data of the inspection vehicle; S4: the plurality of groups of target intersection coordinates of each inspection target are clustered, and a center point position of each inspection target is calculated; Step S4 is specifically as follows: S41: the target intersection coordinates of the inspection target q are divided into k groups, and p target intersection coordinates are randomly selected as clustering centers, and the value of p is less than k; S42: the distance between each target intersection coordinate and each clustering center is calculated, and each target intersection coordinate is assigned to the nearest clustering center; S43: the clustering center and the assigned target intersection coordinate form a cluster; S44: for a cluster, the clustering center is reset according to the target intersection coordinates in the current cluster when a target intersection coordinate is assigned; S45: steps S41-S44 are repeated until the clustering centers in each cluster no longer change; S46: Set the weight according to the validity of the depth of field value, the total number of target intersection coordinates of the inspection target q is N, and the maximum depth of field value is Dep max For target intersection coordinates (U j ,V j ), the weight of the jth target intersection coordinate is: T j = N * Dep max / Dep j where Dep j is the depth of field value of the jth target intersection coordinate, and the maximum value of j is N; S47: Calculate the total weight number of the obtained inspection target q: T = T1 + T2 +... + T N ; S48: the center point position (U, V) of the inspection target q is calculated, and the calculation formula is as follows: U = (U1*T1 + U2*T1 +... + U N *T N ) / T; V = (V1*T1 + V2*T1 +... + V N *T N ) / T.

2. The target positioning correction method for on-vehicle patrol inspection according to claim 1, characterized by, Step S1 is specifically as follows: S11: Obtain the angles A1, A2, …, A between a plurality of track points of the inspection vehicle m and the heading angles B1, B2, …, B of the GPS m ; S12: the installation error angle of the GPS is calculated, and the calculation formula is as follows: C m = (A m -B m +720) % 360 C av = (C 1+ C2+...+C m ) / m wherein m represents the m groups of data collected, C av is the installation error angle of the GPS, and % represents the remainder. S13: after the installation error angle is corrected, the included angle between the direction of the inspection target and the north direction is calibrated to obtain the north direction angle.

3. The method of claim 1, wherein the method is used for a vehicle-mounted target positioning and correction system. Step S2 is specifically as follows: S21: the camera on the inspection vehicle collects an image at a preset frequency; S22: the identification region of each inspection target in the image is obtained, and the included angle a between the straight line formed by the center point of the identification region S and the camera and the driving direction is obtained; S23: in the continuous frame collection, if a i a i-1 and S i >S i-1 , it is judged as the same target, i is the number of the collected frame; when multiple inspection targets appear at the same time, multiple inspection targets are distinguished according to the distance from the depth of field L, the distance from the depth of field L is the distance of the target image pixel point from the camera; S24: each inspection target is respectively tracked to obtain a plurality of groups of original GPS data of the inspection targets.

4. The method of claim 1, wherein, Step S3 is specifically as follows: S31: Extract two adjacent track points of the inspection vehicle in the GPS track data of the inspection vehicle, the coordinates of the track point Point(j) are (X j ,Y j ), the driving direction is R j , and the angle of the binocular recognition to the inspection target q is W j ; the coordinates of the track point Point(j+1) are (X j+1 ,Y j+1 ), the driving direction is R j+1 , and the angle of the binocular recognition to the inspection target q is W j+1 ; j is the number of the track point of the inspection vehicle, and q is the number of the inspection target; S32: the first angle angle1 and the second angle angle2 of the inspection target q are calculated, and the calculation formula is as follows: angle1 = (R j +W j +720) % 360 angle2 = (R j+1 +W j+1 +720) % 360; S33: Calculate the jth target intersection point coordinate (U j ,V j ) of the inspection target q through the first angle angle1 and the second angle angle2, and the calculation formula is: S34: steps S31-S33 are repeated to obtain all target intersection coordinates of the inspection target q.

5. A target positioning correction system for vehicle-mounted inspection, characterized by comprising: a target positioning correction device according to any one of claims 1 to 4; and a vehicle-mounted inspection device. The steps for implementing the method in any one of claims 1-4 comprise: a north direction angle acquisition module, configured to cause the inspection vehicle to inspect according to a preset route, obtain an installation error angle of a GPS, and calculate a north direction angle through the installation error angle; a target tracking module, configured to cause a camera on the inspection vehicle to respectively track a plurality of inspection targets through a target tracking algorithm, and obtain a plurality of groups of original GPS data of the inspection targets; a target intersection coordinate acquisition module, configured to calculate a plurality of groups of target intersection coordinates of each inspection target through the plurality of groups of original GPS data of each inspection target, the north direction angle and GPS track data of the inspection vehicle; a center point position acquisition module, configured to cluster the plurality of groups of target intersection coordinates of each inspection target, and calculate a center point position of each inspection target.

Images