A license plate position recognition method and system, and computer storage medium

Abstract

The present invention discloses a license plate position recognition method and system, and a computer storage medium, including: obtaining a depth image taken by a vehicle-mounted depth camera; performing point cloud conversion on the depth image to obtain a first point cloud; filtering the first point cloud according to a preset filtering rule to obtain a second point cloud; downsampling the second point cloud to obtain a third point cloud; performing point cloud segmentation on the third point cloud to obtain multiple point cloud clusters; performing feature extraction on the multiple point cloud clusters to obtain multiple point cloud feature descriptors; obtaining a preset license plate feature descriptor, and performing feature matching on the multiple point cloud feature descriptors with the license plate feature descriptors, wherein the point cloud corresponding to the successfully matched point cloud feature descriptor is the license plate point cloud, and the position information of the license plate point cloud is determined as the license plate position information. Through the present invention, the technical defects of the existing visual image recognition method and the special material sensing recognition method can be overcome.

Description

Technical Field

[0001] The present invention relates to the technical field of license plate position recognition, and in particular to a license plate position recognition method and system, and a computer-readable storage medium. Background Art

[0002] License plate recognition is a part of automobile intelligence such as assisted driving, automatic driving or parking management; license plate recognition includes the recognition of the license plate position and the recognition of the license plate number characters. The present invention is aimed at the recognition of the license plate position; currently, the recognition of the license plate position is mainly completed by vision-based image recognition or sensing recognition methods based on special materials.

[0003] Among them, the vision-based two-dimensional image recognition method is the most common approach. It segments the image, usually by edge segmentation, semantic segmentation, etc., and then classifies and recognizes the segmented graphics. There are simple graphics processing algorithms, machine learning algorithms, and deep learning methods to achieve the final license plate target sorting; but it is based on two-dimensional image recognition and cannot obtain three-dimensional information, so the license plate recognition accuracy is low;

[0004] Among them, the sensing and recognition method based on special materials mainly uses a terminal carried in the design and production of the license plate itself. The terminal contains special materials to facilitate the capture of traffic cameras; but its scene adaptability is poor. When the license plate is a metal with too high or too low surface reflectivity, its detection effect is poor. Summary of the invention

[0005] The purpose of the present invention is to propose a license plate position recognition method and system, and a computer-readable storage medium to overcome the above-mentioned technical defects of the existing vision-based two-dimensional image recognition method and the sensing recognition method based on special materials.

[0006] To achieve the above object, the present invention provides a method for recognizing a license plate position, comprising:

[0007] Get the depth image taken by the vehicle-mounted depth camera;

[0008] Performing point cloud conversion on the depth image to obtain a first point cloud;

[0009] Filtering the first point cloud according to a preset filtering rule to obtain a second point cloud;

[0010] Downsampling the second point cloud to obtain a third point cloud;

[0011] Performing point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters;

[0012] Performing feature extraction on the multiple point cloud clusters respectively to obtain multiple point cloud feature descriptors;

[0013] A preset license plate feature descriptor is obtained, and the multiple point cloud feature descriptors are respectively matched with the license plate feature descriptor, wherein the point cloud corresponding to the successfully matched point cloud feature descriptor is the license plate point cloud, and the position information of the license plate point cloud is determined as the license plate position information.

[0014] Optionally, acquiring a depth image captured by a vehicle-mounted depth camera includes:

[0015] The exposure time parameter of the vehicle-mounted depth camera is adaptively adjusted so that the depth image captured by the vehicle-mounted depth camera can filter out the vehicle body contour, thereby highlighting the license plate contour.

[0016] Optionally, filtering the first point cloud according to a preset filtering rule to obtain a second point cloud includes:

[0017] Performing a filtering operation to filter out points in the first point cloud whose depth values ​​are equal to a preset depth value;

[0018] A secondary filtering is performed to filter out outlier points in the first point cloud.

[0019] Optionally, downsampling the second point cloud to obtain a third point cloud includes:

[0020] The second point cloud is divided into a plurality of cubes, and one point is retained in each cube, thereby obtaining a third point cloud.

[0021] Optionally, the performing point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters includes:

[0022] The distances between points in the third point cloud are calculated according to a preset algorithm, and the third point cloud is segmented into a plurality of point cloud clusters according to the calculated distances between points and a preset segmentation principle.

[0023] Optionally, the extracting features of the plurality of point cloud clusters respectively to obtain a plurality of point cloud feature descriptors comprises:

[0024] Find the maximum and minimum coordinates of all points in each point cloud cluster;

[0025] Calculate the maximum and minimum points of each point cloud cluster and calculate the length and width of each point cloud cluster;

[0026] A point cloud feature descriptor of each point cloud cluster is obtained according to the length and width of each point cloud cluster, and the point cloud feature descriptor includes the diagonal length and diagonal angle of the point cloud cluster.

[0027] Optionally, the license plate feature descriptor includes a preset diagonal length reference value and a diagonal angle reference value;

[0028] The performing feature matching on the plurality of point cloud feature descriptors and the license plate feature descriptors respectively includes:

[0029] When the error between the diagonal length of any point cloud feature descriptor and the diagonal length reference value is less than a first preset value, and the error between the diagonal angle and the diagonal angle reference value is less than a second preset value, the point cloud feature descriptor is successfully matched with the license plate feature descriptor.

[0030] Optionally, the method further comprises:

[0031] First display information, second display information, third display information and fourth display information are generated respectively according to the first point cloud, second point cloud, third point cloud and license plate point cloud, and the first display information, second display information, third display information and fourth display information are sent to the vehicle display unit for synchronous display.

[0032] A second aspect of the present invention provides a license plate position recognition system, which is used to implement the license plate position recognition method according to the first aspect, and the system comprises:

[0033] An image acquisition unit, used to acquire a depth image taken by a vehicle-mounted depth camera;

[0034] A point cloud conversion unit, configured to perform point cloud conversion on the depth image to obtain a first point cloud;

[0035] A point cloud filtering unit, configured to filter the first point cloud according to a preset filtering rule to obtain a second point cloud;

[0036] A point cloud sampling unit, configured to downsample the second point cloud to obtain a third point cloud;

[0037] A point cloud segmentation unit, used for performing point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters;

[0038] A feature extraction unit, configured to extract features from the plurality of point cloud clusters respectively to obtain a plurality of point cloud feature descriptors; and

[0039] The feature matching unit is used to obtain a preset license plate feature descriptor, and perform feature matching on the multiple point cloud feature descriptors and the license plate feature descriptor respectively, wherein the point cloud corresponding to the successfully matched point cloud feature descriptor is the license plate point cloud, and the position information of the license plate point cloud is determined as the license plate position information.

[0040] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the license plate position recognition method described in the first aspect are implemented.

[0041] In summary, various aspects of the present invention respectively propose a license plate position recognition method and system, and a computer-readable storage medium. In the above aspects, the license plates of surrounding vehicles are photographed by a vehicle-mounted depth camera to obtain a depth image, the photographed depth image is transformed into a point cloud, and filtered and segmented in sequence to obtain multiple point cloud clusters, and the multiple point cloud clusters are subjected to feature extraction to obtain multiple feature descriptors, and the multiple feature descriptors are matched with the preset license plate feature descriptors to determine whether they are license plate point clouds, and finally the position information is extracted from the successfully matched point cloud information as the license plate position information, and the position of the empty parking space is accurately identified within the effective distance. Compared with the traditional two-dimensional image recognition method based on vision, the embodiment of the present invention can obtain three-dimensional stereo information, greatly improving the recognition accuracy in the license plate depth; and compared with the existing sensing recognition method based on special materials, the embodiment of the present invention can adaptively detect the license plate position without ensuring the integrity of the outer contour of the metal with too high or too low reflectivity on the vehicle surface, thereby improving the robustness of the license plate position detection.

[0042] Other features and advantages of the present invention will be set forth in the description which follows. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0044] Figure 1 The present invention is a flowchart of a license plate position recognition method in one embodiment of the present invention.

[0045] Figure 2 The figure is a schematic diagram of the structure of a license plate position recognition system in another embodiment of the present invention. DETAILED DESCRIPTION

[0046] Various exemplary embodiments, features and aspects of the present disclosure will be described in detail below with reference to the accompanying drawings. In addition, in order to better illustrate the present invention, numerous specific details are given in the specific embodiments below. It should be understood by those skilled in the art that the present invention can also be implemented without certain specific details. In some examples, means well known to those skilled in the art are not described in detail in order to highlight the subject matter of the present invention.

[0047] See also Figure 1 An embodiment of the present invention provides a license plate position recognition method, comprising the following steps S1-S7:

[0048] Step S1, obtaining a depth image taken by a vehicle-mounted depth camera;

[0049] Exemplarily, the step S1 may include:

[0050] Adaptively adjusting the exposure time parameter of the vehicle-mounted depth camera so that the depth image captured by the vehicle-mounted depth camera can filter out the vehicle body contour, thereby highlighting the license plate contour;

[0051] Specifically, the depth camera used in this embodiment is preferably but not limited to a depth camera based on the TOF principle. The shooting of the depth map relies on the reflection of the laser. The integrity of metals with too high or too low surface reflectivity cannot be guaranteed. Under fixed conditions, adjusting the exposure time of the depth camera can receive more or less reflected light. Prolonging the exposure time can receive more reflected light, making the outline of the photographed object more complete; shortening the exposure time can reduce the received reflected light, making the outline of the photographed object incomplete and highlighting the highlight part; this embodiment uses the physical property that the license plate has a high reflectivity to light, and deliberately shortens the exposure time of the depth camera to filter out the outline of the vehicle body and highlight the outline of the license plate.

[0052] Step S2, performing point cloud conversion on the depth image to obtain a first point cloud;

[0053] Specifically, the pixel of the depth camera is, for example, 320×240. According to the internal parameters of the depth camera, the depth image is transformed into a three-dimensional coordinate. The first point cloud obtained after the transformation contains 76,800 points. All points are represented by three-dimensional space coordinates (X, Y, Z). The origin of the reference coordinate of the transformation is the optical center of the depth camera, and the reference coordinate system is established according to the right-hand rule.

[0054] Step S3, filtering the first point cloud according to a preset filtering rule to obtain a second point cloud;

[0055] Specifically, the first point cloud obtained in step S2 is an original point cloud, which may also contain some invalid points or outliers, and needs to be filtered to obtain a second point cloud;

[0056] Exemplarily, the step S3 includes:

[0057] Step S31, performing a filtering operation to filter out points in the first point cloud whose depth values ​​are equal to a preset depth value;

[0058] Specifically, the primary filtering is conditional filtering. For example, the effective detection range of the depth camera used is a maximum of 6.0 meters. For points beyond the detection range, the depth value is considered to be the preset maximum depth. The depth value corresponds to Z in the three-dimensional space coordinates (X, Y, Z). Therefore, points where Z is equal to the maximum depth are considered invalid points and need to be filtered out through conditional filtering. In this embodiment, the conditional filtering sets points where Z is greater than 0 and less than 5.9 as valid points;

[0059] Step S32: performing secondary filtering to filter out outliers in the first point cloud;

[0060] Specifically, the secondary filtering is statistical filtering, that is, outliers are identified by counting the number of neighboring points of all points in the point cloud, and the outliers are treated as noise points for elimination;

[0061] Preferably, in this embodiment, the number of neighboring points of the query point considered by the statistical filter during statistics is 50, and the threshold for determining whether it is an outlier point is 0.3;

[0062] Step S4, downsampling the second point cloud to obtain a third point cloud;

[0063] Exemplarily, the step S4 includes:

[0064] Divide the second point cloud into a plurality of cubes, retaining one point in each cube, thereby obtaining a third point cloud;

[0065] Specifically, in this example, the point cloud is downsampled using voxel grids, that is, the point cloud is divided into many small cubes, and one point is retained in each cube to reduce the number of points without destroying the distribution characteristics of the point cloud, thereby reducing the computational complexity of post-processing. In this example, the voxel block is set to a cube of 0.01×0.01×0.01 meters;

[0066] Step S5, performing point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters;

[0067] Exemplarily, step S5 may include:

[0068] Calculate the distances between points in the third point cloud according to a preset algorithm, and divide the third point cloud into a plurality of point cloud clusters according to the calculated distances between points and a preset segmentation principle;

[0069] Specifically, the preset algorithm is, for example, a Euclidean clustering segmentation algorithm. The calculated distance is the Euclidean distance between points. The Euclidean clustering segmentation algorithm can be used to segment the point cloud into several point cloud clusters based on the principle of proximity. In this embodiment, the search radius of the nearest neighbor search in the Euclidean clustering segmentation is set to 0.02 meters, the minimum number of points required for a clustered point cloud cluster is set to 100, and the maximum number of points is set to 25,000. The point cloud search mechanism can use the Kd tree method;

[0070] Step S6, performing feature extraction on the multiple point cloud clusters respectively to obtain multiple point cloud feature descriptors;

[0071] In this embodiment, a feature descriptor is used to describe the features of the point cloud cluster. Exemplarily, step S6 may include:

[0072] Step S61, finding the maximum value and minimum value of the coordinates of all points in each point cloud cluster;

[0073] Specifically, the points in each point cloud cluster are traversed to find the maximum value point max_p and the minimum value point min_p of the (X, Y, Z) coordinates of all points in the point cloud cluster; the maximum value points max_p and min_p represent the two diagonal points of the lower left corner and the upper right corner of the point cloud cluster respectively;

[0074] Step S62, calculating the maximum value point and the minimum value point of each point cloud cluster and calculating the length and width of each point cloud cluster;

[0075] Specifically, as described above, the maximum value points max_p and min_p represent the two diagonal points of the lower left corner and the upper right corner of the point cloud group, respectively. Therefore, the length and width of the point cloud group can be calculated according to the coordinate parameters of the maximum value point max_p and the minimum value point min_p obtained in step S61;

[0076] deta_x = max_p.x - min_p.x;

[0077] deta_y = max_p.y - min_p.y;

[0078] Among them, deta_x is the length, deta_y is the width, max_p.x represents the horizontal coordinate of the maximum value point, max_p.y represents the vertical coordinate of the maximum value point, min_p.x represents the horizontal coordinate of the minimum value point, and min_p.y represents the vertical coordinate of the minimum value point;

[0079] Step S63, obtaining a point cloud feature descriptor of each point cloud cluster according to the length and width of each point cloud cluster, wherein the point cloud feature descriptor includes a diagonal length and a diagonal angle of the point cloud cluster;

[0080]

[0081] angle=tan -1 deta_y / deta_x;

[0082] Where distance is the length of the diagonal line, and angle is the angle of the diagonal line.

[0083] Step S7, obtaining a preset license plate feature descriptor, performing feature matching on the multiple point cloud feature descriptors and the license plate feature descriptor respectively, wherein the point cloud corresponding to the successfully matched point cloud feature descriptor is the license plate point cloud, and determining the position information of the license plate point cloud as the license plate position information;

[0084] Exemplarily, the license plate feature descriptor includes a preset diagonal length reference value and a diagonal angle reference value;

[0085] Correspondingly, performing feature matching on the plurality of point cloud feature descriptors and the license plate feature descriptors respectively includes:

[0086] When the error Δdistance between the diagonal length of any point cloud feature descriptor and the diagonal length reference value is less than a first preset value, and the error Δangle between the diagonal angle and the diagonal angle reference value is less than a second preset value, the point cloud feature descriptor is successfully matched with the license plate feature descriptor;

[0087] Wherein, the first preset value is preferably but not limited to being set to 0.35;

[0088] The second preset value is preferably but not limited to being set to 0.32.

[0089] In a specific embodiment, the method further comprises:

[0090] Step S8, generate first display information, second display information, third display information, and fourth display information according to the first point cloud, the second point cloud, the third point cloud, and the license plate point cloud, respectively, and send the first display information, the second display information, the third display information, and the fourth display information to the vehicle display unit for synchronous display.

[0091] Specifically, point cloud visualization is proposed in this embodiment. The point cloud visualization first creates a window, and then creates four viewpoints in the window of the vehicle display unit. Each viewpoint is displayed in a linked manner, that is, when a viewpoint is dragged to change the rotation angle or zoom in and out, the other three viewpoints also undergo the same change, that is, they are displayed synchronously; the four viewpoints are respectively the first display information, the second display information, the third display information, and the fourth display information.

[0092] Based on the above description, it can be known that the method of the embodiment of the present invention uses the vehicle-mounted depth camera to shoot the license plates of surrounding vehicles to obtain a depth image, transforms the captured depth image into a point cloud, and sequentially filters and segments to obtain multiple point cloud clusters, and extracts features from the multiple point cloud clusters to obtain multiple feature descriptors, and matches the multiple feature descriptors with the preset license plate feature descriptors to determine whether they are license plate point clouds, and finally extracts the position information from the successfully matched point cloud information as the license plate position information, and accurately identifies the position of the empty parking space within the effective distance. Compared with the traditional vision-based two-dimensional image recognition method, the embodiment of the present invention can obtain three-dimensional stereo information, greatly improving the recognition accuracy of the license plate depth; and compared with the existing sensing and recognition method based on special materials, the embodiment of the present invention does not need to ensure the integrity of the outer contour of the metal with too high or too low reflectivity on the vehicle surface, and can also adaptively detect the license plate position, thereby improving the robustness of the license plate position detection.

[0093] See also Figure 2 Another embodiment of the present invention provides a license plate position recognition system. The system of this embodiment corresponds to the method described in the above embodiment and can be used to implement the method steps of the above embodiment. The system of this embodiment specifically includes:

[0094] An image acquisition unit 1 is used to acquire a depth image taken by a vehicle-mounted depth camera;

[0095] A point cloud conversion unit 2, configured to perform point cloud conversion on the depth image to obtain a first point cloud;

[0096] A point cloud filtering unit 3, configured to filter the first point cloud according to a preset filtering rule to obtain a second point cloud;

[0097] A point cloud sampling unit 4, configured to downsample the second point cloud to obtain a third point cloud;

[0098] A point cloud segmentation unit 5, configured to perform point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters;

[0099] A feature extraction unit 6, configured to extract features from the plurality of point cloud clusters respectively to obtain a plurality of point cloud feature descriptors; and

[0100] The feature matching unit 7 is used to obtain a preset license plate feature descriptor, and perform feature matching on the multiple point cloud feature descriptors and the license plate feature descriptor respectively, wherein the point cloud corresponding to the successfully matched point cloud feature descriptor is the license plate point cloud, and the position information of the license plate point cloud is determined as the license plate position information.

[0101] The system embodiment described above is merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0102] It should be noted that the system described in the above embodiment corresponds to the method described in the above embodiment. Therefore, the undescribed part of the system described in the above embodiment can be obtained by referring to the content of the method described in the above embodiment, that is, the specific step content of steps S1-S7 of the method in the above embodiment can be understood as the functions that can be achieved by the system of this embodiment, and will not be repeated here.

[0103] Furthermore, if the license plate position recognition system described in the above embodiment is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium.

[0104] Another embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the license plate position recognition method described in the above embodiment are implemented.

[0105] Specifically, the computer-readable storage medium may include: any entity or device that can carry the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (R value OM, Read-Only MemoR value y), random access memory (R value AM, Random Access MemoR value y), electrical carrier signal, telecommunication signal and software distribution medium, etc.

[0106] The embodiments of the present invention have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.

Claims

1. A method for recognizing a license plate position, characterized in that: include: Get the depth image taken by the vehicle-mounted depth camera; Performing point cloud conversion on the depth image to obtain a first point cloud; Filtering the first point cloud according to a preset filtering rule to obtain a second point cloud; Downsampling the second point cloud to obtain a third point cloud; Performing point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters; Extracting features from the plurality of point cloud clusters respectively to obtain a plurality of point cloud feature descriptors, including: finding the maximum value point and the minimum value point of the coordinates of all points in each point cloud cluster; Calculate the maximum and minimum points of each point cloud cluster and the length and width of each point cloud cluster; obtain the point cloud feature descriptor of each point cloud cluster according to the length and width of each point cloud cluster, wherein the point cloud feature descriptor includes the diagonal length and diagonal angle of the point cloud cluster; A preset license plate feature descriptor is obtained, wherein the license plate feature descriptor includes a preset diagonal length reference value and a diagonal angle reference value, and feature matching is performed on the multiple point cloud feature descriptors respectively with the license plate feature descriptor, wherein the point cloud corresponding to the successfully matched point cloud feature descriptor is the license plate point cloud, and the position information of the license plate point cloud is determined as the license plate position information.

2. The license plate position recognition method according to claim 1, characterized in that: The step of obtaining a depth image captured by a vehicle-mounted depth camera includes: The exposure time parameter of the vehicle-mounted depth camera is adaptively adjusted so that the depth image captured by the vehicle-mounted depth camera can filter out the vehicle body contour, thereby highlighting the license plate contour.

3. The license plate position recognition method according to claim 1, characterized in that: The filtering the first point cloud according to a preset filtering rule to obtain a second point cloud includes: Performing a filtering operation to filter out points in the first point cloud whose depth values ​​are equal to a preset depth value; A secondary filtering is performed to filter out outlier points in the first point cloud.

4. The license plate position recognition method according to claim 1, characterized in that: The downsampling the second point cloud to obtain a third point cloud includes: The second point cloud is divided into a plurality of cubes, and one point is retained in each cube, thereby obtaining a third point cloud.

5. The license plate position recognition method according to claim 1, characterized in that: The step of performing point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters includes: The distances between points in the third point cloud are calculated according to a preset algorithm, and the third point cloud is segmented into a plurality of point cloud clusters according to the calculated distances between points and a preset segmentation principle.

6. The license plate position recognition method according to claim 1, characterized in that: The performing feature matching on the plurality of point cloud feature descriptors and the license plate feature descriptors respectively includes: When the error between the diagonal length of any point cloud feature descriptor and the diagonal length reference value is less than a first preset value, and the error between the diagonal angle and the diagonal angle reference value is less than a second preset value, the point cloud feature descriptor is successfully matched with the license plate feature descriptor.

7. The license plate position recognition method according to claim 1, characterized in that: The method further comprises: First display information, second display information, third display information and fourth display information are generated respectively according to the first point cloud, second point cloud, third point cloud and license plate point cloud, and the first display information, second display information, third display information and fourth display information are sent to the vehicle display unit for synchronous display.

8. A license plate position recognition system, characterized in that: For implementing the license plate position recognition method according to any one of claims 1 to 7, the system comprises: An image acquisition unit, used to acquire a depth image taken by a vehicle-mounted depth camera; A point cloud conversion unit, configured to perform point cloud conversion on the depth image to obtain a first point cloud; A point cloud filtering unit, configured to filter the first point cloud according to a preset filtering rule to obtain a second point cloud; A point cloud sampling unit, configured to downsample the second point cloud to obtain a third point cloud; A point cloud segmentation unit, used for performing point cloud segmentation on the third point cloud to obtain a plurality of point cloud clusters; A feature extraction unit, configured to extract features from the plurality of point cloud clusters respectively to obtain a plurality of point cloud feature descriptors; and The feature matching unit is used to obtain a preset license plate feature descriptor, and perform feature matching on the multiple point cloud feature descriptors and the license plate feature descriptor respectively, wherein the point cloud corresponding to the successfully matched point cloud feature descriptor is the license plate point cloud, and the position information of the license plate point cloud is determined as the license plate position information.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the license plate position recognition method described in any one of claims 1 to 7 are implemented.

Images