Method and apparatus for obtaining point cloud data

Abstract

The application provides a kind of point cloud data acquisition method and device, by acquiring road image data, at least one pixel point is selected as target pixel point from the image pixel range covered by road sign, the laser point cluster corresponding to target pixel point is obtained according to the laser point cloud collected when collecting road image, based on the laser point cluster corresponding to target pixel point, the laser point located in the preset range around laser point cluster is obtained from laser point cloud, the laser point belonging to road sign is identified by the shape type of laser point, laser point cluster and road sign, the problem that prior art cannot quickly identify the point cloud data belonging to road sign from the point cloud data of road image is solved, the interference of other factors on the road is avoided, the efficiency and accuracy of identification are improved, and important basic data is provided for the production of road sign.

Description

Technical Field

[0001] This invention relates to the field of high-precision mapping, and more specifically, to a method and apparatus for acquiring point cloud data. Background Technology

[0002] Electronic maps have become deeply integrated into people's lives, and their functions are becoming increasingly sophisticated. Currently, people's needs for electronic maps have gone beyond traditional functions such as location positioning and route guidance. Especially with the rise of autonomous driving technology, the development of high-precision map technology, as the data foundation for autonomous driving, plays a crucial role. Autonomous vehicles lack the inherent visual and logical abilities of human drivers, while high-precision maps rely on precise three-dimensional representations of road networks and contain a wealth of information that can be used for driving assistance, such as intersection layouts and road sign locations. This information is the basis for judging road conditions in autonomous driving.

[0003] Currently, a crucial element of high-precision maps is road signs. Representing road signs in high-precision maps requires not only conveying the content on the signs but also their spatial location. Obtaining this spatial location relies on point cloud data. Existing technology typically involves manually searching for point cloud data belonging to road signs, a method that suffers from low efficiency. Therefore, finding the point cloud data belonging to road signs quickly is a technical problem that needs to be solved. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a method and apparatus for acquiring point cloud data, which solves the problem that the prior art cannot quickly identify point cloud data belonging to road signs from road point cloud data. By using image and point cloud matching and recognition, point cloud data belonging to road signs can be quickly acquired, avoiding interference from other factors on the road and improving the efficiency and accuracy of recognition.

[0005] To solve the above-mentioned technical problems, the following solution is proposed:

[0006] A method for acquiring point cloud data includes:

[0007] Acquire road image data, which includes the image pixel range covered by road signs in the road image and the shape type of the road signs;

[0008] From the image pixel range covered by the road sign, at least one pixel is selected as the target pixel;

[0009] Based on the laser point cloud acquired during the acquisition of the road image, obtain the laser point cluster corresponding to the target pixel;

[0010] Based on the laser point cluster corresponding to the target pixel, obtain the laser points located within a preset range around the laser point cluster from the laser point cloud;

[0011] The laser points belonging to the road signs are identified by the shape and type of the laser points, laser point clusters, and road signs.

[0012] Preferably, acquiring road image data specifically includes:

[0013] The system acquires road images captured by the camera device of the data acquisition vehicle, and then trains the road images using a pre-established fast convolutional neural network training model to obtain the image pixel range covered by the road signs in the road images and the shape type of the road signs.

[0014] Preferably, at least one pixel is selected as the target pixel from the image pixel range covered by the road sign, specifically including:

[0015] Select the center pixel of the image pixel range covered by the road sign as the target pixel.

[0016] Preferably, the laser point cluster corresponding to the target pixel is obtained based on the laser point cloud acquired during the acquisition of the road image, specifically including:

[0017] Obtain the projection matrix and attitude information of the laser that generated the laser point cloud;

[0018] Obtain the intrinsic and extrinsic parameter matrix of the camera device of the data acquisition vehicle;

[0019] Using the projection matrix, attitude information, and intrinsic and extrinsic parameter matrices, a transformation matrix between the laser point cloud and the road image is established;

[0020] Based on the transformation matrix, the laser point cluster corresponding to the target pixel is obtained from the laser point cloud.

[0021] Preferably, based on the laser point cluster corresponding to the target pixel, laser points located within a preset range around the laser point cluster are obtained from the laser point cloud, specifically including:

[0022] Using the laser point cluster corresponding to the target pixel as a reference point, laser points located within a preset range around the laser point cluster are obtained from the laser point cloud through clustering.

[0023] A point cloud data acquisition device, comprising:

[0024] An image processing unit is used to acquire road image data, the road image data including the image pixel range covered by road signs in the road image and the shape type of the road signs, and to select at least one pixel point as a target pixel point from the image pixel range covered by the road signs.

[0025] The point cloud processing unit is used to obtain the laser point cluster corresponding to the target pixel based on the laser point cloud collected when the road image is acquired, and to obtain laser points located within a preset range around the laser point cluster from the laser point cloud based on the laser point cluster corresponding to the target pixel.

[0026] The point cloud acquisition unit is used to identify the laser points belonging to the road sign based on the shape and type of the laser points, laser point clusters, and road signs.

[0027] Preferably, the image processing unit acquires road image data, specifically including:

[0028] The system acquires road images captured by the camera device of the data acquisition vehicle, and then trains the road images using a pre-established fast convolutional neural network training model to obtain the image pixel range covered by the road signs in the road images and the shape type of the road signs.

[0029] Preferably, the image processing unit selects at least one pixel as the target pixel from the image pixel range covered by the road sign, specifically including:

[0030] Select the center pixel point located at the geometric center of the image pixel range covered by the road sign as the target pixel point.

[0031] Preferably, the point cloud processing unit obtains the laser point cluster corresponding to the target pixel based on the laser point cloud acquired during the acquisition of the road image, specifically including:

[0032] Obtain the projection matrix and attitude information of the laser that generated the laser point cloud;

[0033] Obtain the intrinsic and extrinsic parameter matrix of the camera device of the data acquisition vehicle;

[0034] Using the projection matrix, attitude information, and intrinsic and extrinsic parameter matrices, a transformation matrix between the laser point cloud and the road image is established;

[0035] Based on the transformation matrix, the laser point cluster corresponding to the target pixel is obtained from the laser point cloud.

[0036] Preferably, the point cloud processing unit, based on the laser point cluster corresponding to the target pixel, obtains laser points located within a preset range around the laser point cluster from the laser point cloud, specifically including:

[0037] Using the laser point cluster corresponding to the target pixel as a reference point, laser points located within a preset range around the laser point cluster are obtained from the laser point cloud through clustering.

[0038] A point cloud data acquisition device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the steps of a point cloud data acquisition method.

[0039] A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of a method for acquiring point cloud data.

[0040] As can be seen from the above technical solutions, the point cloud data acquisition method provided in this application acquires road image data, selects at least one pixel as the target pixel from the image pixel range covered by the road sign, acquires the laser point cloud collected when acquiring the road image, obtains the laser point cluster corresponding to the target pixel, and acquires the laser points located within a preset range around the laser point cluster from the laser point cloud based on the laser point cluster corresponding to the target pixel. By identifying the laser points, laser point clusters and the shape type of the road sign, the laser points belonging to the road sign are identified. This solves the problem that the prior art cannot quickly identify the point cloud data belonging to the road sign from the point cloud data of the road, thereby effectively avoiding interference from other factors on the road, improving the efficiency and accuracy of identification, and providing important basic data for the production of road signs. Attached Figure Description

[0041] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0042] Figure 1 This is a flowchart of the point cloud data acquisition method of the present invention.

[0043] Figure 2 This is a schematic diagram of the projection matrix of the camera in the point cloud data acquisition method of the present invention.

[0044] Figure 3 This is a schematic diagram of the mapping between the camera's projection matrix and the laser point cloud matrix in the point cloud data acquisition method of the present invention.

[0045] Figure 4This is a schematic diagram of the point cloud data acquisition method based on geometric shape type recognition in the present invention.

[0046] Figure 5 This is one of the structural schematic diagrams of the point cloud data acquisition device of the present invention.

[0047] Figure 6 This is the second schematic diagram of the point cloud data acquisition device of the present invention.

[0048] Figure 7 This is a schematic diagram of the point cloud data acquisition system of the present invention. Detailed Implementation

[0049] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0050] The point cloud data acquisition method provided by this invention is applicable to the field of electronic maps, especially to the field of high-precision electronic maps involving autonomous driving and advanced driver assistance systems, and is used for processing auxiliary information such as road signs in high-precision maps.

[0051] High-precision maps, also known as high-resolution maps, are characterized primarily by their accuracy. If the positional difference between the geographic features depicted on a high-precision map and their counterparts in the real world is within the centimeter level, then the map's accuracy is considered centimeter-level. Since autonomous vehicles lack the inherent visual and logical abilities of human drivers, accuracy is crucial for ensuring their safety. Therefore, high-precision maps contain a wealth of driver assistance information, the most important of which is a precise three-dimensional representation of the road network, such as intersection layouts and road sign locations. They also include a great deal of semantic information, such as the starting position of left-turn lanes and speed limit indicators.

[0052] Point cloud is a collection of point data on the surface of an object obtained by measuring instruments. Points with large distances between them are called sparse point clouds. High-precision maps usually use dense point clouds, which are obtained by scanning the road surface and surrounding environment (such as road signs, lampposts, etc.) with LiDAR. These dense point clouds have a large number of points and are relatively dense.

[0053] like Figure 1 As shown, the present invention provides a method for acquiring point cloud data, which specifically includes:

[0054] Step 101: Obtain road image data, which includes the image pixel range covered by road signs in the road image and the shape type of the road signs.

[0055] Road images acquired by data acquisition devices form the most basic data foundation for creating high-precision maps. Typically, high-precision data acquisition vehicles (hereinafter referred to as acquisition vehicles) are used for road image acquisition. These vehicles are equipped with cameras capable of capturing road images, lidar capable of acquiring point cloud arrays, and positioning equipment capable of tracking the vehicle's trajectory. Trajectory points are location points collected from the positioning points output by the positioning equipment as the vehicle travels on the road. Acquisition can be performed at certain time periods or at certain frequencies, and the coordinates of the trajectory points are generally latitude and longitude coordinates.

[0056] LiDAR (LiDAR) generates laser point clouds based on laser measurement principles. Each laser point typically includes a timestamp and / or reflection intensity information. The timestamp can be the moment each laser point in the point cloud was acquired, or it can be the reflection duration of the laser point, i.e., the time between the LiDAR emitting the laser beam and receiving the laser point reflected back from that beam.

[0057] Reflection intensity is the echo intensity collected by the receiving device of a lidar system. Reflection intensity is related to the surface material and roughness of the target, the angle of incidence, the emission energy of the instrument, and the laser wavelength. Reflection intensity can be used to determine whether there are obstructions on the road or objects of the same or similar medium. For example, if a point cloud matrix with high reflection intensity appears in a road with low reflection intensity, it is generally considered to be a specific object.

[0058] By combining the parameters of the laser point cloud collected by lidar (including at least reflection intensity and reflection duration) with the coordinates based on trajectory points, three-dimensional coordinates can be generated for each laser point in the laser point cloud, that is, three-dimensional coordinates including latitude, longitude and altitude coordinates, such as Ri eGL engineering coordinates.

[0059] A camera device captures road images by taking photos or videos. The road image data obtained in step 101 of this invention is road image data obtained after manual or machine recognition of road images collected by a high-precision map acquisition vehicle. This includes the image pixel range covered by road signs and the shape type of the road signs. Since a single road image / frame may contain more than just road signs, the image pixel range covered by road signs in this application refers to the range of pixels covered by the road signs in the road image. Furthermore, the shape type of the road signs includes circular, square, and rhomboid shapes, with the specific shape depending on the shape of the road signs captured in the road image.

[0060] Step 102: Select at least one pixel from the image pixel range covered by the road sign as the target pixel.

[0061] Within the pixel range covered by the road sign, one or more pixels are arbitrarily selected as the target pixel. To reduce the difficulty of subsequent point cloud expansion, the preferred method provided by this invention includes: calculating the geometric center of the pixel range covered by the road sign based on its geometric shape, obtaining the center pixel of the geometric shape of the pixel range covered by the road sign as the center pixel of the road sign obtained from the road image data, and using this center pixel as the target pixel.

[0062] Step 103: Obtain the laser point clusters corresponding to the target pixels based on the laser point cloud collected when acquiring road images.

[0063] Based on the target pixel in the road image data, find the corresponding laser point cluster in the laser point cloud data collected at the same time.

[0064] Because the camera and lidar are not installed at completely identical positions and angles on the data acquisition vehicle, the road image data captured by the camera and the point cloud array do not completely overlap in space. In other words, the road signs in the road image data and the road sign point clouds in the point cloud array cannot be directly correlated. It is necessary to obtain the approximate position based on the attitude information of the camera and then make further adjustments to obtain the position of the road sign point cloud.

[0065] Specifically, the projection matrix and attitude information of the laser that generated the laser point cloud are obtained, the intrinsic and extrinsic parameter matrices of the camera device of the acquisition vehicle are obtained, and a transformation matrix between the laser point cloud and the road image is established using the projection matrix, attitude information, and intrinsic and extrinsic parameter matrices. Based on the transformation matrix, the laser point clusters corresponding to the target pixel are obtained from the laser point cloud (there is more than one laser point corresponding to the target pixel).

[0066] like Figure 3 As shown, the coordinates of the camera device and the image it captures are obtained. The image captured by the camera device is quantized into a projection matrix. The position of the target pixel of the road sign in the projection matrix is ​​found. The shooting posture and angle are calculated based on the position of the target pixel and the line connecting the camera device.

[0067] like Figure 4 As shown, the position of the camera device is simulated in the laser point cloud array, or it can be said that the above projection matrix is ​​replaced by the laser point cloud array. A cluster of rays is emitted from the position of the camera device according to the calculated shooting posture and angle. The point where the cluster of rays intersects with the point cloud array is considered to be the area where the road sign is located.

[0068] The images and point cloud arrays used in this invention can be collected synchronously by the acquisition vehicle during the same acquisition process on the same road; or they can be collected by the acquisition vehicle at basically similar positions and in basically similar postures during different acquisition processes on the same road. The point cloud data collected in different acquisition processes can be fused together before the solution of this invention is executed.

[0069] When using the center pixel of the pixel range covered by a road sign in road image data as the target pixel, it's important to note that because the camera and LiDAR are installed in different positions on the data acquisition vehicle, the laser points mapped from the center pixel of the road sign in the captured image to the laser point cluster in the point cloud array may not all belong to the center point of the road sign in the point cloud data. However, due to the size limitations of the data acquisition vehicle and the fact that the camera and LiDAR are usually not installed too far apart during data acquisition, the laser point cluster mapped from the center pixel of the road sign in the captured image is highly likely to contain the laser points included in the road sign in the point cloud array. Therefore, a laser point cloud belonging to the road sign can be obtained based on this laser point cluster.

[0070] Step 104: Based on the laser point cluster corresponding to the target pixel, obtain the laser points located within a preset range around the laser point cluster from the laser point cloud.

[0071] As mentioned above, the laser point cluster determined in step 103 must be part of the point cloud data of the road sign. Based on the position of the laser point cluster, a certain range is extended outward to obtain the laser points within the surrounding preset range.

[0072] Based on the geometric shape type, the laser point cloud matrix within a preset range can be obtained by radiating outwards at a preset distance from the laser point cluster corresponding to the target pixel.

[0073] Specifically, a preset distance range (in centimeters) is used to extend the laser point cluster corresponding to the central pixel outwards. The selection of this preset distance range is generally based on the standard size of the road sign and the scale of the point cloud array relative to the real object. There are many ways to extend it outwards. For example, with coordinate point P as the geometric center, it can radiate outwards by 10 cm, 20 cm, 50 cm, 100 cm, and 200 cm, obtaining the coordinate range of the entire radiation range as the range of the laser point cloud matrix of the sign.

[0074] The preset range is usually chosen to be relatively large, but the distance parameter can also be changed to meet different situations, which will be explained in detail in the following steps.

[0075] Furthermore, the expanded laser point cloud matrix is ​​clustered to remove obvious noise points. There are many ways to cluster laser points; this invention can use reflection intensity clustering or density clustering, as detailed below.

[0076] As mentioned earlier, if a point cloud matrix with high reflection intensity appears within a road point cloud matrix with low reflection intensity, it is generally considered that the point cloud matrix with high reflection intensity represents a specific object. The reflection intensity of the laser points in the expanded point cloud matrix is ​​clustered, and laser points with reflection intensity higher than a preset threshold are selected as the point cloud representation of the road sign.

[0077] If the expanded point cloud matrix contains many laser points with reflection intensities exceeding the preset threshold, it may be due to an insufficiently large selected range or an off-center location of the laser point cluster corresponding to the target pixel, resulting in an incomplete representation of the road sign's laser points in the current point cloud array. In this case, it is necessary to further expand the preset range and then re-cluster the point cloud reflection intensities.

[0078] Clustering can also be done using density clustering, such as the BSCAN density clustering algorithm. These density clustering algorithms generally assume that clusters can be determined by the density of the sample distribution. Samples of the same cluster are closely connected; that is, any sample in that cluster will have other samples of the same cluster nearby. By grouping closely connected samples into one cluster, we obtain a cluster category. By dividing all closely connected groups into different categories, we obtain the final clustering results.

[0079] Step 105: Identify the laser points belonging to the road signs by using the laser points, laser point clusters, and the shape type of the road signs.

[0080] Step 104 allows for relatively accurate acquisition of the relevant point cloud data of the road sign. Furthermore, based on the laser point clusters and geometric shape type, the external contour of the acquired road sign laser point cloud data is identified and its accuracy is corrected. This is done because when capturing images of road signs or acquiring point cloud arrays using LiDAR, interference from objects nearby may occur. For example, if there is a tree next to the road sign, or someone is standing next to it, the calculated point cloud of the road sign will be disturbed, and its external contour will appear irregular. Normally, road signs have regular shapes, and it is unreasonable for irregular shapes to appear in the point cloud data.

[0081] To overcome this interference, the type of geometric shape identified from road image data can serve as a basis for further refining the point cloud of road signs. Geometric shape types include common shapes such as squares, rhombuses, triangles, and circles, for example... Figure 4 The image shows an example of a circular outline, the shape of which is determined by the shape of the graphic's border line segments.

[0082] First, the external contour of the laser point cloud data of the road sign is identified to obtain its external contour information and determine its geometric shape type. Then, it is matched with the geometric shape type identified from the road image data. If the types match, it means that the determined point cloud of the road sign is relatively accurate and does not need to be modified. If the types do not match, it means that the determined point cloud of the road sign is not accurate enough and there is interference, and the external contour of the point cloud of the road sign needs to be corrected.

[0083] There are several methods to correct the outer contour of the point cloud of a road sign. For example, one can first find the geometric center of the point cloud matrix of the road sign, and then redetermine the contour edges according to the type of geometry, so that the contour can retain the maximum number of point clouds. Of course, other edge denoising methods commonly used in this field can also be used.

[0084] Based on the same concept as the point cloud data acquisition method provided above, this invention also provides a point cloud data acquisition device, such as... Figure 5 As shown, the device includes: an image processing unit 100, a point cloud processing unit 200, and a point cloud acquisition unit 300.

[0085] The image processing unit 100 is used to acquire road image data, which includes the image pixel range covered by road signs in the road image and the shape type of the road signs. From the image pixel range covered by the road signs, at least one pixel is selected as the target pixel.

[0086] The point cloud processing unit 200 is used to obtain the laser point cluster corresponding to the target pixel based on the laser point cloud acquired when acquiring the road image; and to obtain the laser points located within a preset range around the laser point cluster from the laser point cloud based on the laser point cluster corresponding to the target pixel.

[0087] The point cloud acquisition unit 300 is used to identify the laser points belonging to the road signs based on the laser points, laser point clusters and the shape type of the road signs acquired by the point cloud processing unit 200.

[0088] Based on the same concept as the point cloud data acquisition method provided above, this invention also provides a point cloud data acquisition device, such as... Figure 6As shown, the device includes a memory 101, a processor 102, and a computer program stored in the memory and executable on the processor 102. When the processor 102 executes the computer program, it implements the steps of a method for acquiring point cloud data.

[0089] Based on the same concept as the point cloud data acquisition method provided above, this invention also provides a point cloud data acquisition system, such as... Figure 7 As shown, the system includes a camera device 1000, a laser point cloud device (such as a lidar sensor) 2000, and a point cloud generation device 3000.

[0090] Both the camera device 1000 and the laser point cloud device 2000 are installed on the data acquisition vehicle to collect data synchronously while driving on the road. The camera device 1000 collects road images, and the laser point cloud device 2000 collects point cloud arrays.

[0091] The point cloud generation device 3000 acquires and stores the road image collected by the camera device 1000, the attitude information of the camera device 1000, and the point cloud array collected by the laser point cloud device 2000.

[0092] The point cloud generation device 3000 includes: an image processing unit 100, a point cloud processing unit 200, and a point cloud acquisition unit 300.

[0093] The image processing unit 100 is used to acquire road image data, which includes the image pixel range covered by road signs in the road image and the shape type of the road signs. From the image pixel range covered by the road signs, at least one pixel is selected as the target pixel.

[0094] The point cloud processing unit 200 is used to obtain the laser point cluster corresponding to the target pixel based on the laser point cloud acquired when acquiring the road image; and to obtain the laser points located within a preset range around the laser point cluster from the laser point cloud based on the laser point cluster corresponding to the target pixel.

[0095] The point cloud acquisition unit 300 is used to identify the laser points belonging to the road signs based on the laser points, laser point clusters and the shape type of the road signs acquired by the point cloud processing unit 200.

[0096] Finally, 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 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 limitation, 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.

[0097] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For apparatus embodiments, since they are fundamentally similar to method embodiments, the descriptions are relatively simple; relevant parts can be referred to the descriptions in the method embodiments.

[0098] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0099] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0100] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0101] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. 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 application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

[0102] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of methods, apparatus, and computer program products according to various embodiments of the present invention. In this regard, each block in the flowcharts and block diagrams may represent a module, program segment, or portion of code, containing one or more computer-executable instructions for implementing logical functions. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. Furthermore, it should be noted that each block or combination of blocks in the block diagrams and flowcharts may be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

Claims

1. A method for acquiring point cloud data, the method comprising: Acquire road image data, which includes the image pixel range covered by road signs in the road image and the shape type of the road signs; From the image pixel range covered by the road sign, at least one pixel is selected as the target pixel; Based on the laser point cloud acquired during the acquisition of the road image, obtain the laser point cluster corresponding to the target pixel; Based on the laser point cluster corresponding to the target pixel, obtain the laser points located within a preset range around the laser point cluster from the laser point cloud; The laser points belonging to the road signs are identified by the shape and type of the laser points, laser point clusters, and road signs.

2. The method according to claim 1, specifically including: acquiring road image data, comprising: The system acquires road images captured by the camera device of the data acquisition vehicle, and then trains the road images using a pre-established fast convolutional neural network training model to obtain the image pixel range covered by the road signs in the road images and the shape type of the road signs.

3. The method according to claim 1, wherein at least one pixel is selected as the target pixel from the image pixel range covered by the road sign, specifically comprising: Select the center pixel of the image pixel range covered by the road sign as the target pixel.

4. The method according to claim 1, wherein obtaining the laser point cluster corresponding to the target pixel based on the laser point cloud acquired during the acquisition of the road image, specifically includes: Obtain the projection matrix and attitude information of the laser that generated the laser point cloud; Obtain the intrinsic and extrinsic parameter matrix of the camera device on the data acquisition vehicle; Using the projection matrix, attitude information, and intrinsic and extrinsic parameter matrices, a transformation matrix between the laser point cloud and the road image is established; Based on the transformation matrix, the laser point cluster corresponding to the target pixel is obtained from the laser point cloud.

5. The method according to claim 1, wherein, based on the laser point cluster corresponding to the target pixel, laser points located within a preset range surrounding the laser point cluster are obtained from the laser point cloud, specifically including: Using the laser point cluster corresponding to the target pixel as a reference point, laser points located within a preset range around the laser point cluster are obtained from the laser point cloud through clustering.

6. A point cloud data acquisition device, the device comprising: An image processing unit is used to acquire road image data, the road image data including the image pixel range covered by road signs in the road image and the shape type of the road signs, and to select at least one pixel point as a target pixel point from the image pixel range covered by the road signs. The point cloud processing unit is used to obtain the laser point cluster corresponding to the target pixel based on the laser point cloud collected when the road image is acquired, and to obtain laser points located within a preset range around the laser point cluster from the laser point cloud based on the laser point cluster corresponding to the target pixel. The point cloud acquisition unit is used to identify the laser points belonging to the road sign based on the shape and type of the laser points, laser point clusters, and road signs.

7. The apparatus according to claim 6, wherein the image processing unit acquires road image data, specifically comprising: The system acquires road images captured by the camera device of the data acquisition vehicle, and then trains the road images using a pre-established fast convolutional neural network training model to obtain the image pixel range covered by the road signs in the road images and the shape type of the road signs.

8. The apparatus according to claim 6, wherein the image processing unit selects at least one pixel as the target pixel from the image pixel range covered by the road sign, specifically comprising: Select the center pixel of the image pixel range covered by the road sign as the target pixel.

9. The apparatus according to claim 6, wherein the point cloud processing unit obtains the laser point cluster corresponding to the target pixel based on the laser point cloud acquired during the acquisition of the road image, specifically comprising: Obtain the projection matrix and attitude information of the laser that generated the laser point cloud; Obtain the intrinsic and extrinsic parameter matrix of the camera device on the data acquisition vehicle; Using the projection matrix, attitude information, and intrinsic and extrinsic parameter matrices, a transformation matrix between the laser point cloud and the road image is established; Based on the transformation matrix, the laser point cluster corresponding to the target pixel is obtained from the laser point cloud.

10. The apparatus according to claim 6, wherein the point cloud processing unit, based on the laser point cluster corresponding to the target pixel, obtains laser points located within a preset range surrounding the laser point cluster from the laser point cloud, specifically including: Using the laser point cluster corresponding to the target pixel as a reference point, laser points located within a preset range around the laser point cluster are obtained from the laser point cloud through clustering.

11. An apparatus comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the computer program, implements the steps of the method as claimed in any one of claims 1-5.

12. A computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method as claimed in any one of claims 1-5.

Images