Sign three-dimensional reconstruction method, device and equipment and storage medium

Abstract

The present disclosure relates to a road sign three-dimensional reconstruction method, device, equipment and storage medium. The method comprises: acquiring a road image sequence; performing road sign recognition processing and road sign tracking processing in the road image sequence to obtain a bounding box of the same road sign in each road image; based on the bounding box, performing feature point matching processing on the road sign in at least part of the road images to obtain the pixel positions of the points on the road sign in different road images; based on the pixel positions of the points on the road sign in different road images, determining a space plane where the road sign is located; and mapping the pixels of the road sign in at least one road image to the space plane to obtain a three-dimensional point cloud of the road sign. The present disclosure can improve the accuracy of three-dimensional point cloud data by identifying and tracking the road sign in the road image sequence, performing feature point matching, mapping the pixels of the road sign in the image to the road sign plane determined according to the points, and obtaining the three-dimensional point cloud of the road sign.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a method, apparatus, device and storage medium for three-dimensional reconstruction of road signs. Background Technology

[0002] Autonomous driving is a scenario where control is achieved through a computer system without human intervention. In this scenario, the 3D reconstruction of road signs is crucial for route planning, road recognition, and other autonomous driving operations. Therefore, how to perform 3D reconstruction of road signs and improve the accuracy of this reconstruction is a key technical problem that needs to be solved. Summary of the Invention

[0003] To solve the above-mentioned technical problems, or at least partially solve them, this disclosure provides a method, apparatus, device, and computer-readable storage medium for three-dimensional reconstruction of road signs.

[0004] In a first aspect, embodiments of this disclosure provide a method for three-dimensional reconstruction of road signs, including:

[0005] Obtain road image sequences;

[0006] In the road image sequence, road sign recognition and tracking processes are performed to obtain the bounding box of the same road sign in each road image;

[0007] Based on the bounding box, feature point matching processing is performed on the road sign in at least some road images to obtain the pixel positions of the points on the road sign in different road images;

[0008] The spatial plane in which the road sign is located is determined based on the pixel positions of the points on the road sign in different road images;

[0009] The pixels of the road sign in at least one road image are mapped onto the spatial plane to obtain the three-dimensional point cloud of the road sign.

[0010] Secondly, embodiments of this disclosure provide a road sign three-dimensional reconstruction device, comprising:

[0011] The acquisition module is used to acquire road image sequences;

[0012] The recognition and tracking module is used to perform road sign recognition and tracking processing in the road image sequence to obtain the bounding box of the same road sign in each road image.

[0013] A matching module is used to perform feature point matching processing on the road sign in at least some road images based on the bounding box, so as to obtain the pixel position of the point on the road sign in different road images;

[0014] The determination module is used to determine the spatial plane where the road sign is located based on the pixel positions of the points on the road sign in different road images;

[0015] A mapping module is used to map the pixels of the road sign in at least one road image onto the spatial plane to obtain a three-dimensional point cloud of the road sign.

[0016] Thirdly, embodiments of this disclosure provide a computer device, including:

[0017] Memory;

[0018] Processor; and

[0019] Computer programs;

[0020] The computer program is stored in the memory and configured to be executed by the processor to implement the method as described in the first aspect.

[0021] Fourthly, embodiments of this disclosure provide a computer-readable storage medium having a computer program stored thereon, the computer program being executed by a processor to implement the method described in the first aspect.

[0022] The road sign 3D reconstruction method, apparatus, device, and storage medium provided in this disclosure identify and track road signs in a sequence of obtained road images to obtain bounding boxes of the same road sign in each road image. Based on the bounding boxes, feature point matching is performed on the same road sign in at least some road images to obtain the pixel positions of the points on the road sign in different road images. Based on the pixel positions of the points on the road sign in different road images, the spatial plane where the road sign is located in actual space is determined. Then, the pixels of the road sign in at least one road image are mapped to the spatial plane where the road sign is actually located. This can accurately obtain the 3D point cloud of the road sign and improve the accuracy of road sign 3D reconstruction. Attached Figure Description

[0023] The accompanying drawings, which are incorporated in and form a part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure.

[0024] To more clearly illustrate the technical solutions in the embodiments of this disclosure or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A schematic diagram illustrating an application scenario provided by an embodiment of this disclosure;

[0026] Figure 2 A flowchart illustrating an exemplary method for three-dimensional reconstruction of a road sign, provided as an embodiment of this disclosure;

[0027] Figure 3 A flowchart of a feature point matching method provided in this disclosure embodiment;

[0028] Figure 4 A flowchart of another feature point matching method provided in this disclosure embodiment;

[0029] Figure 5 A schematic diagram of the structure of a three-dimensional reconstruction device for a road sign provided in an embodiment of this disclosure;

[0030] Figure 6 A schematic diagram of the structure of a computer device provided in an embodiment of this disclosure. Detailed Implementation

[0031] To better understand the above-mentioned objectives, features, and advantages of this disclosure, the solutions disclosed herein will be further described below. It should be noted that, unless otherwise specified, the embodiments and features described herein can be combined with each other.

[0032] Numerous specific details are set forth in the following description in order to provide a full understanding of this disclosure, but this disclosure may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only some, and not all, of the embodiments of this disclosure.

[0033] Typically, during autonomous driving, vehicles need to perform operations such as route planning and road sign recognition. These operations require the support of road sign information. Therefore, how to perform 3D reconstruction of road signs and improve the accuracy of the 3D reconstruction is a technical problem that needs to be solved. To address this problem, this disclosure provides a method for 3D reconstruction of road signs, which will be described below with reference to specific embodiments.

[0034] Figure 1 This is a schematic diagram illustrating an application scenario provided by an embodiment of this disclosure. The application scenario includes an autonomous vehicle 11, a road sign 12, and a camera 13. The method provided by this embodiment can be used for... Figure 1 The road sign 12 shown is subjected to 3D reconstruction.

[0035] Example, Figure 2 This is a flowchart illustrating an exemplary method for 3D reconstruction of road signs, provided as an embodiment of this disclosure. The method can be executed by a 3D reconstruction device, which can be implemented in software and / or hardware. This 3D reconstruction device can be understood as a computer device or a functional module within a computer device. The computer device includes, but is not limited to, a vehicle's infotainment system or a drone's controller.Figure 2 As shown, the method may include the following steps:

[0036] S201. Obtain the road image sequence.

[0037] The road image sequence referred to in this disclosure can be understood as a sequence of road images acquired in continuous time, which includes multiple consecutive road images.

[0038] The road image sequence in this embodiment can be a real-time acquired image sequence or a pre-acquired image sequence. For example, in Figure 1 In the scenario shown, a camera mounted on a vehicle can continuously capture road images, and the real-time captured road image sequence can be sent to the 3D reconstruction device described in this embodiment of the disclosure, so that the 3D reconstruction device can perform 3D reconstruction of the road surface elements in the road image sequence. Alternatively, in other embodiments, the road image sequence captured by the camera can also be stored in a preset database, and the 3D reconstruction device described in this embodiment of the disclosure can obtain the pre-captured road image sequence from the database and perform 3D reconstruction processing.

[0039] S202. Perform road sign recognition and tracking processing on the road image sequence to obtain the bounding box of the same road sign in each road image.

[0040] The bounding box referred to in the embodiments of this disclosure can be understood as the smallest outer bounding box of the road sign obtained after recognizing the road sign in the road image. The shape of the bounding box can be preset as needed, and the commonly used bounding box shapes are circles and rectangles.

[0041] In this embodiment of the disclosure, road sign recognition processing can be performed on each road image in a road image sequence to obtain the bounding box of the road sign on each road image. A road image may include images of one or more road signs; when a road image includes images of multiple road signs, the bounding boxes of multiple road signs can be obtained through image recognition processing.

[0042] In this embodiment of the disclosure, there are various methods for tracking road image sequences. For example, feature similarity can be calculated by comparing features to determine the feature similarity of images within different bounding boxes, and road signs in two bounding boxes with a feature similarity higher than a preset similarity can be identified as the same road sign. Alternatively, the overlap rate between bounding boxes on adjacent road images can be calculated, and road signs in two bounding boxes with an overlap rate higher than a preset threshold can be identified as the same road sign.

[0043] In this embodiment of the disclosure, determining the overlap rate between bounding boxes on adjacent road images involves selecting two adjacent road images and converting them into images in the same coordinate system. This coordinate system can be the coordinate system of either of the two images or a third coordinate system different from the coordinate systems of the two images. The positions of the road sign bounding boxes in the two converted images are compared to obtain the ratio of the area of ​​the overlapping part of the two bounding boxes to the area of ​​the covered part of the two bounding boxes. This yields the overlap rate between the bounding boxes on adjacent road images. When the overlap rate is higher than a preset threshold, the bounding boxes of the road signs in the two images can be considered as the bounding boxes of the same road sign.

[0044] S203. Based on the bounding box, perform feature point matching processing on the road sign in at least some road images to obtain the pixel positions of the points on the road sign in different road images.

[0045] The "at least part of the road images" mentioned in this disclosure can be understood as all the road images in a road image sequence, or as a part of the road images in a road image sequence. For example, if a road image sequence contains 50 road images, feature point matching can be performed among 30 of them.

[0046] The feature points mentioned in this disclosure can be understood as points with stable features on a road sign image, where the features remain unchanged when the image changes.

[0047] The feature point matching mentioned in this embodiment can be understood as performing feature matching on feature points on different road sign images, and determining two feature points with a feature similarity higher than a preset threshold as the same feature point. The same feature point corresponds to the same point on the road sign in reality.

[0048] For example, in some implementations, the Scale Invariant Feature Transform (SIFT) algorithm can be used to describe the feature points detected in the road sign image using a 128-dimensional feature vector. A road sign image after SIFT is represented as a 128-dimensional feature vector set, which has features invariant to scaling, translation, and rotation of the road sign image. After scale-space extremum detection, feature points are accurately located, their orientation is determined, and feature vectors are generated, achieving feature point matching. Methods for feature point matching based on the SIFT algorithm can be found in related technologies and will not be elaborated upon here. Of course, the SIFT algorithm is only one available feature point matching method, not the only one. In practice, different feature matching methods can be selected as needed to achieve feature point matching for road signs, such as optical flow methods and Oriented Fast and Rotated Brief (ORB) algorithms.

[0049] S204. Based on the pixel positions of the points on the road sign in different road images, determine the spatial plane where the road sign is located.

[0050] The spatial plane referred to in this disclosure can be understood as a two-dimensional plane in three-dimensional space, and the spatial plane where the road sign is located can be understood as the two-dimensional plane in which the road sign is located in actual space.

[0051] In one exemplary embodiment of this disclosure, the spatial plane where the road sign is located can be determined based on the principle that a point on a plane and the direction of the plane's normal can uniquely determine a plane. Specifically, the coordinates of the point closest to the shooting device on the spatial plane where the road sign is located can be determined based on the pixel position of the point on the road sign in different road images; then, the direction of the line connecting the target point and the shooting device is taken as the normal direction of the plane, thereby determining the spatial plane where the road sign is located based on the coordinates of the target point and the direction of the line connecting the target point and the shooting device.

[0052] In this embodiment of the disclosure, the coordinates of the target point can be determined by obtaining the pixel coordinates of each pair of matching points and the camera pose when capturing the road image. Based on the calculation method of the line-plane intersection and the camera intrinsic parameter matrix as known parameters, a residual function for each pair of matching points can be constructed. The residual can be understood as the Euclidean distance on the plane after the two matching points are projected onto the road sign plane. The residual function can be expressed as follows:

[0053]

[0054] Among them, f i The meaning is the residual of the i-th pair of points, p cThe meaning is the target point coordinates, used as parameters to represent the plane. P represents the specific projection function, which can be calculated based on the intersection of the line and the plane. K is the camera intrinsic parameter matrix. i A Let T be the pixel coordinates of the i-th matching point pair in frame A. i A The camera pose of frame A corresponding to the i-th matching point pair, p i B Let T be the pixel coordinates of the i-th matching point pair in frame B. i B The camera pose of the corresponding B-frame for the i-th set of matching points.

[0055] The residuals of all matching points together constitute the final objective function:

[0056]

[0057] The coordinates of the target point are variables of the objective function. The objective function is solved by nonlinear optimization, and the coordinates of the point that minimizes the objective function value are the coordinates of the target point.

[0058] S205. Map the pixels of the road sign in at least one road image onto the spatial plane to obtain the three-dimensional point cloud of the road sign.

[0059] The "at least one road image" mentioned in the embodiments of this disclosure can be any of all road images, or one or more road images from all road images. For example, in one embodiment, for a specific road sign, the road image with the largest bounding box size corresponding to the road sign can be determined from the road image sequence as the target road image. The pixels of the road sign in the mv bag road image are then mapped to the spatial plane where the road sign is located to obtain the three-dimensional point cloud of the road sign.

[0060] This embodiment of the disclosure identifies and tracks road signs in the obtained road image sequence to obtain bounding boxes of the same road sign in each image. Feature point matching is performed on some or all of the images based on the bounding boxes. The spatial plane where the road sign is located is determined based on the pixel positions of the points on the road sign in different images. At least one road image is selected, and its pixels are mapped to the spatial plane where the road sign is located. The resulting point set is the three-dimensional point cloud of the road sign, which can improve the accuracy of the three-dimensional point cloud data.

[0061] Figure 3 This is a flowchart illustrating a feature point matching method provided in an embodiment of this disclosure. Figure 3 As shown in the above embodiments, feature point matching processing of road signs on different road images can be performed using the following method.

[0062] S301. Based on the shooting positions of each road image obtained in advance, determine a number of road images from the road image sequence whose shooting distance from each other is less than a preset distance.

[0063] In this embodiment of the disclosure, the shooting location can be understood as the location of the vehicle when shooting the road image. For example, the shooting location can be obtained by high-precision GPS on the vehicle, or the accurate positioning information of the vehicle can be obtained based on methods such as Simultaneous Localization and Mapping (SLAM). For specific execution methods, please refer to related technologies, which will not be elaborated here.

[0064] In this embodiment of the disclosure, the shooting distance can be understood as the distance between the shooting positions of two road images, which can be calculated based on the shooting positions of the two road images.

[0065] In this embodiment of the disclosure, multiple road images captured at a distance less than a preset distance can be understood as multiple road images captured in low-speed driving or parking scenarios.

[0066] S302, Delete one or more of the plurality of road images from the road image sequence.

[0067] When a vehicle is traveling at low speed or parked, the camera on the vehicle still takes pictures at a fixed shooting frequency. In this case, since the distance the vehicle moves in a unit of time is very small, many of the road images captured by the camera are repetitive. By judging the shooting distance between the road images, one or more images that are close to each other can be deleted, so that the remaining images have a large parallax, improving the accuracy of road sign reconstruction, while reducing the amount of data and increasing the processing speed.

[0068] S303. Perform feature point matching on the road sign in the remaining road images.

[0069] In this embodiment of the disclosure, the same feature points are matched in the remaining image sequence after deletion. The specific method can be referred to the technical solution in S203, which will not be repeated here.

[0070] This embodiment of the disclosure can improve the accuracy of road sign reconstruction by deleting some road images collected in low-speed or parking scenarios, thereby increasing the parallax between the remaining images, reducing the amount of data, and improving the processing speed.

[0071] Figure 4 A flowchart illustrating another feature point matching method provided in this embodiment of the disclosure. Figure 5As shown in the above embodiments, feature point matching processing of road signs on different road images can be performed using the following method.

[0072] S401. Based on the size of the bounding box of the road sign on each road image, the road images with a higher bounding box than the preset order are selected as road images to be processed, in descending order of size.

[0073] In this embodiment of the present disclosure, the bounding boxes of road signs in the obtained road image sequence are sorted in descending order. The area covered by the bounding box can be used as the basis for judging the size of the bounding box, and road images with a higher sorting order than the preset sorting order are selected as images to be processed.

[0074] S402. Perform feature point matching on the road sign in the road image to be processed.

[0075] In this embodiment of the disclosure, the same feature points are matched in the remaining image sequence after deletion. The specific method can be referred to the technical solution in S203, which will not be repeated here.

[0076] This embodiment of the disclosure sorts the image sequence according to the bounding box size, selects road images with a higher sorting order than the preset sorting, and performs feature point matching processing on the road signs in the images. This can reduce the impact of excessively small parallax on the subsequent acquisition of planar parameters, and at the same time, it can extract more feature points from the road sign images, further improving the accuracy of the 3D point cloud data.

[0077] It is important to note that Figure 3 and Figure 4 The technical solutions in the two embodiments shown can be used individually, or the technical solutions of the two embodiments can be combined and applied to the three-dimensional reconstruction device of road signs.

[0078] Figure 5 This is a schematic diagram of the structure of the three-dimensional reconstruction device for road signs provided in an embodiment of this disclosure. Figure 5As shown, the 3D reconstruction device 50 includes: an acquisition module 51, an identification and tracking module 52, a matching module 53, a determination module 54, and a mapping module 55; wherein, the acquisition module 51 is used to acquire a sequence of road images; the identification and tracking module 52 is used to perform road sign identification and tracking processing in the road image sequence to obtain bounding boxes of the same road sign in each road image; the matching module 53 is used to perform feature point matching processing on the road sign in at least some road images based on the bounding boxes to obtain the pixel positions of the points on the road sign in different road images; the determination module 54 is used to determine the spatial plane where the road sign is located based on the pixel positions of the points on the road sign in different road images; and the mapping module 55 is used to map the pixels of the road sign in at least one road image onto the spatial plane to obtain a 3D point cloud of the road sign.

[0079] Optionally, the recognition and tracking module 52 includes: a recognition unit 521 and a tracking unit 522; the recognition unit 521 is used to perform road sign recognition processing on each road image to obtain the bounding box of the road sign on each road image; the tracking unit 522 is used to determine the overlap rate between the bounding boxes on adjacent road images, and to determine the bounding boxes with an overlap rate higher than a preset threshold as the bounding boxes of the same road sign.

[0080] Optionally, the matching module 53 is used to: determine multiple road images from the road image sequence whose shooting distance from each other is less than a preset distance based on the shooting positions of each road image obtained in advance; delete one or more of the multiple road images from the road image sequence; and perform feature point matching on the road sign in the remaining road images.

[0081] Optionally, the matching module 53 is used to: based on the size of the bounding box of the road sign on each road image, and in descending order of the bounding box size, select road images with a higher ranking than a preset ranking as road images to be processed; and perform feature point matching on the road sign in the road images to be processed.

[0082] Optionally, the determining module 54 is used to: determine the coordinates of a target point based on the pixel position of the point on the road sign in different road images, wherein the target point refers to the point on the spatial plane that is closest to the shooting device; and determine the spatial plane based on the coordinates of the target point and the direction of the line connecting the target point and the shooting device.

[0083] Optionally, the mapping module 55 is used to: for the road sign, determine the road image with the largest bounding box size from the road image road column as the target road image; map the pixels of the road sign in the target road image onto the spatial plane to obtain the three-dimensional point cloud of the road sign.

[0084] Figure 5 The three-dimensional reconstruction device for road signs in the illustrated embodiment can be used to execute the technical solutions of the above method embodiments. Its implementation principle and technical effects are similar, and will not be repeated here.

[0085] Figure 6 This is a schematic diagram of the structure of a computer device provided in an embodiment of this disclosure. The computer device provided in this embodiment of the disclosure can process the methods described in the above embodiments, such as... Figure 6 As shown, the computer device 60 includes: a memory 61, a processor 62, a computer program, and a communication interface 63; wherein the computer program is stored in the memory 61 and configured to be executed by the processor 62 as described above in the road sign 3D reconstruction method.

[0086] In addition, this disclosure also provides a computer-readable storage medium storing a computer program thereon, which is executed by a processor to implement the road sign 3D reconstruction method described in the above embodiments.

[0087] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus 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 apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0088] The above description is merely a specific embodiment of this disclosure, enabling those skilled in the art to understand or implement it. 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 this disclosure. Therefore, this disclosure is not to be limited to the embodiments described herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for three-dimensional reconstruction of road signs, characterized in that, The method includes: Obtain road image sequences; In the road image sequence, road sign recognition and tracking processes are performed to obtain the bounding box of the same road sign in each road image; Based on the bounding box, feature point matching processing is performed on the road sign in at least some road images to obtain the pixel positions of the points on the road sign in different road images; The spatial plane in which the road sign is located is determined based on the pixel positions of the points on the road sign in different road images; The pixels of the road sign in at least one road image are mapped onto the spatial plane to obtain the three-dimensional point cloud of the road sign; Determining the spatial plane where the road sign is located based on the pixel positions of points on the road sign in different road images includes: Based on the pixel positions of the points on the road sign in different road images, the coordinates of the target point are determined. The target point refers to the point on the spatial plane that is closest to the shooting device. The spatial plane is determined based on the coordinates of the target point and the direction of the line connecting the target point and the shooting device.

2. The method according to claim 1, characterized in that, The process of performing road sign recognition and tracking in the road image sequence to obtain the bounding box of the same road sign in each road image includes: Perform road sign recognition processing on each road image to obtain the bounding box of the road sign on each road image; Determine the overlap rate between bounding boxes on adjacent road images, and identify bounding boxes with an overlap rate higher than a preset threshold as bounding boxes of the same road sign.

3. The method according to claim 1, characterized in that, The step of performing feature point matching processing on the road sign in at least a portion of the road image based on the bounding box includes: Based on the shooting positions of each road image obtained in advance, a number of road images with a shooting distance of less than a preset distance are determined from the road image sequence; Delete one or more of the plurality of road images from the road image sequence; Feature point matching is performed on the road signs in the remaining road images.

4. The method according to claim 1, characterized in that, The step of performing feature point matching processing on the road sign in at least a portion of the road image based on the bounding box includes: Based on the size of the bounding box of the road sign on each road image, road images with a higher bounding box than the preset order are selected as road images to be processed, in descending order of size. Feature point matching is performed on the road sign in the road image to be processed.

5. The method according to claim 1, characterized in that, The step of mapping the pixels of the road sign in at least one road image onto the spatial plane to obtain the three-dimensional point cloud of the road sign includes: For the road sign, the road image with the largest bounding box size is determined from the road image sequence as the target road image; The pixels of the road sign in the target road image are mapped onto the spatial plane to obtain the three-dimensional point cloud of the road sign.

6. A three-dimensional reconstruction device for road signs, characterized in that, include: The acquisition module is used to acquire road image sequences; The recognition and tracking module is used to perform road sign recognition and tracking processing in the road image sequence to obtain the bounding box of the same road sign in each road image. A matching module is used to perform feature point matching processing on the road sign in at least some road images based on the bounding box, so as to obtain the pixel position of the point on the road sign in different road images; The determination module is used to determine the spatial plane where the road sign is located based on the pixel positions of the points on the road sign in different road images; A mapping module is used to map the pixels of the road sign in at least one road image onto the spatial plane to obtain a three-dimensional point cloud of the road sign; The determination module is used for: Based on the pixel positions of the points on the road sign in different road images, the coordinates of the target point are determined. The target point refers to the point on the spatial plane that is closest to the shooting device. The spatial plane is determined based on the coordinates of the target point and the direction of the line connecting the target point and the shooting device.

7. The apparatus according to claim 6, characterized in that, The identification and tracking module includes: an identification unit and a tracking unit; The recognition unit is used to perform road sign recognition processing on each road image to obtain the bounding box of the road sign on each road image; The tracking unit is used to determine the overlap rate between bounding boxes on adjacent road images, and to identify bounding boxes with an overlap rate higher than a preset threshold as bounding boxes of the same road sign.

8. The apparatus according to claim 6, characterized in that, The matching module is used for: Based on the shooting positions of each road image obtained in advance, a number of road images with a shooting distance of less than a preset distance are determined from the road image sequence; Delete one or more of the plurality of road images from the road image sequence; Feature point matching is performed on the road signs in the remaining road images.

9. The apparatus according to claim 6, characterized in that, The matching module is used for: Based on the size of the bounding box of the road sign on each road image, road images with a higher bounding box than the preset order are selected as road images to be processed, in descending order of size. Feature point matching is performed on the road sign in the road image to be processed.

10. The apparatus according to claim 6, characterized in that, The mapping module is used for: For the road sign, the road image with the largest bounding box size is determined from the road image road column as the target road image; The pixels of the road sign in the target road image are mapped onto the spatial plane to obtain the three-dimensional point cloud of the road sign.

11. A computer device, characterized in that, include: Memory; processor; as well as Computer programs; The computer program is stored in the memory and configured to be executed by the processor to implement the method as described in any one of claims 1-5.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-5.

Images