Data processing method, device and equipment for point cloud labeling frame welt in automatic driving and storage medium

Abstract

The invention discloses a data processing method, device and equipment for point cloud labeling frame welt in automatic driving and a storage medium. The data processing method comprises the following steps: identifying an actual labeling frame of point cloud data; expanding the actual annotation box to generate a simulated annotation box; based on the point cloud data, obtaining points falling into the simulation annotation box; and calculating the number of the points falling into the simulation labeling box, and when the number of the points is greater than a preset number threshold value, measuring and calculating the fitting degree of the point cloud data and the azimuth plane of the actual labeling box. According to the method, the actual labeling frame of the point cloud data is automatically recognized, the simulation labeling frame is generated through expansion, the number and the fitting degree of the points falling into the simulation frame are calculated based on the point cloud data, and automation and objectification of labeling quality inspection are achieved. According to the method, the labeling efficiency is improved, the dependence on professional labeling personnel is reduced, the labor cost is reduced, the stability of the labeling quality is enhanced, and the interference of human factors is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of autonomous driving, and particularly relates to a data processing method, device, equipment and storage medium for the edge attachment of point cloud annotation frames in autonomous driving. Background Art

[0002] In the field of autonomous driving technology, point cloud data annotation is a key link in realizing environmental perception and target detection. Point cloud data annotation provides training data for tasks such as object detection, semantic segmentation, and instance segmentation in the autonomous driving system by annotating the three-dimensional point cloud data collected by lidar. However, in the process of point cloud fusion annotation, the degree of fit between the ground truth of the annotation result and the actual point cloud is an important indicator to measure the annotation quality.

[0003] It can be seen that it is particularly important to check the degree of fit between the ground truth of the annotation result and the point cloud. The traditional inspection methods mainly rely on manual judgment and adjustment, and there are the following problems:

[0004] (1) Subjectivity: Manual inspection depends on the experience and subjective judgment of annotators, and it is difficult to reach an agreement on the judgment criteria for the same data by different personnel.

[0005] (2) Low efficiency: Manual inspection requires a large amount of time and manpower. Especially when facing large-scale data annotation tasks, efficiency becomes a bottleneck.

[0006] (3) High cost: Manual inspection requires a professional annotation team, with high labor costs, and it is difficult to ensure the stability of annotation quality.

[0007] With the development of autonomous driving technology, the demand for point cloud data annotation is increasing continuously. How to improve the annotation efficiency and quality has become an urgent problem to be solved. Summary of the Invention

[0008] In order to overcome the above technical defects, the present invention provides a data processing method, device, equipment and storage medium for the edge attachment of point cloud annotation frames in autonomous driving.

[0009] In order to solve the above problems, the present invention is implemented according to the following technical solutions:

[0010] In a first aspect, the present invention provides a data processing method for the edge attachment of point cloud annotation frames in autonomous driving, and the data processing method includes the following steps:

[0011] Identify the actual annotation frame of the point cloud data;

[0012] Expand the actual annotation frame to generate a simulated annotation frame;

[0013] Based on the point cloud data, obtain the points falling into the simulated annotation frame;

[0014] Calculate the number of points falling within the simulated annotation box. When the number of points is greater than the preset quantity threshold, measure the degree of fit between the point cloud data and the orientation planes of the actual annotation box.

[0015] Combined with the first aspect, the present invention also provides a first preferred implementation manner of the first aspect. Specifically, expand the original annotation box to generate a simulated annotation box, specifically:

[0016] The orientation planes of the actual annotation box include the front, back, left, right, top, and bottom surfaces;

[0017] Expand the front, back, left, right, and top surfaces of the actual annotation box outward by 30 cm respectively;

[0018] Shrink the bottom surface of the actual annotation box inward by 50 cm;

[0019] Generate a simulated annotation box.

[0020] Combined with the first aspect, the present invention also provides a second preferred implementation manner of the first aspect. Specifically, the preset quantity threshold is 100. When the number of points is greater than 100, measure the degree of fit between the point cloud data and the orientation planes of the actual annotation box.

[0021] Combined with the first aspect, the present invention also provides a third preferred implementation manner of the first aspect. Specifically, the measurement of the degree of fit between the point cloud data and the orientation planes of the actual annotation box specifically includes:

[0022] Calculate the perpendicular distance from the points within the simulated annotation box to each orientation plane of the actual annotation box;

[0023] Count the number of points whose distance range from the orientation plane is within 30 cm;

[0024] If the number of points with an outer distance within 30 cm is greater than 5, it is determined that the point cloud data does not fit the orientation plane;

[0025] If the number of points with an inner distance within 30 cm is less than 5, it is determined that the point cloud data does not fit the orientation plane;

[0026] If the number of points with an outer distance within 30 cm is less than or equal to 5, and the number of points with an inner distance within 30 cm is greater than or equal to 5, it is determined that the point cloud data fits the orientation plane.

[0027] Combined with the first aspect, the present invention also provides a fourth preferred implementation manner of the first aspect. Specifically, when the number of points is less than the preset quantity threshold, end the process.

[0028] Second aspect, the present invention provides a data processing device for the edge attachment of point cloud annotation boxes in autonomous driving. The data processing device is configured to execute the data processing method for the edge attachment of point cloud annotation boxes in autonomous driving. The data processing device includes:

[0029] An identification module, which is used to identify the actual annotation box of the point cloud data;

[0030] A simulation module, which is used to expand the actual annotation box to generate a simulated annotation box;

[0031] An acquisition module, which is used to acquire the points falling into the simulated annotation box based on the point cloud data;

[0032] A calculation module, which is used to calculate the number of points falling into the simulated annotation box. When the number of points is greater than a preset number threshold, the fitting degree between the point cloud data and the orientation plane of the actual annotation box is measured.

[0033] Combined with the second aspect, the present invention also provides a first preferred implementation manner of the second aspect. Specifically, the simulation module expands the original annotation box to generate a simulated annotation box, specifically as follows:

[0034] The orientation planes of the actual annotation box include the front, rear, left, right, top, and bottom;

[0035] The front, rear, left, right, and top of the actual annotation box are respectively expanded outward by 30 cm;

[0036] The bottom of the actual annotation box is contracted inward by 50 cm;

[0037] Generate a simulated annotation box.

[0038] Combined with the second aspect, the present invention also provides a second preferred implementation manner of the second aspect. Specifically, the preset number threshold is 100. When the number of points is greater than 100, the fitting degree between the point cloud data and the orientation plane of the actual annotation box is measured.

[0039] Third aspect, the present invention also provides an electronic device, which includes:

[0040] At least one processor; and a memory communicatively connected to the at least one processor;

[0041] Wherein, the memory stores a computer program executable by the at least one processor, and the computer program is

[0042] Executed by the at least one processor, so that the at least one processor can execute the data processing method for the edge attachment of point cloud annotation boxes in the first aspect.

[0043] Fourthly, the present invention also provides a computer-readable storage medium storing a computer program,

[0044] wherein the computer program, when executed by a processor, implements the data processing method for the edge attachment of the point cloud annotation box in the first aspect.

[0045] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0046] The present invention provides a data processing method for the edge attachment of the point cloud annotation box in autonomous driving. The data processing method includes the following steps: identifying the actual annotation box of the point cloud data; expanding the actual annotation box to generate a simulated annotation box; obtaining the points falling into the simulated annotation box based on the point cloud data; calculating the number of points falling into the simulated annotation box, and when the number of points is greater than a preset number threshold, measuring the degree of fit between the point cloud data and the azimuth plane of the actual annotation box.

[0047] Traditional methods rely on manual judgment and have problems such as strong subjectivity, low efficiency, high cost, and unstable annotation quality. The present invention solves many problems of the traditional manual inspection of the point cloud annotation quality through an automated algorithm process. The present invention realizes the automation and objectification of the annotation quality inspection by automatically identifying the actual annotation box of the point cloud data, expanding and generating a simulated annotation box, and calculating the number of points falling into the simulated box and the degree of fit based on the point cloud data. This method not only improves the annotation efficiency, reduces the dependence on professional annotators, and lowers the labor cost, but also enhances the stability of the annotation quality and avoids the interference of human factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The following further elaborates on the specific embodiments of the present invention with reference to the drawings, where:

[0049] Figure 1 is the technical flowchart of a data processing method for the edge attachment of the point cloud annotation box in an embodiment of the present invention;

[0050] Figure 2 is the algorithm flowchart of a data processing method for the edge attachment of the point cloud annotation box in an embodiment of the present invention;

[0051] Figure 3 is the module diagram of a data processing device for the edge attachment of the point cloud annotation box in an embodiment of the present invention;

[0052] Figure 4 is the structural schematic diagram of an electronic device in an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0053] The preferred embodiments of the present invention will be described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only for the purpose of illustrating and explaining the present invention, and are not intended to limit the present invention.

[0054] In point cloud fusion annotation, it is necessary to check the degree of fit between the point cloud of the annotation result truth value and the object. The traditional inspection method relies on manual judgment and adjustment. The effect of this method depends on subjective judgment of people, resulting in difficulty in reaching a consensus on the judgment criteria for the same data, and this method has relatively low efficiency and high labor costs.

[0055] To this end, the present invention provides a data processing method for the edge attachment of a point cloud annotation box in autonomous driving. The data processing method includes the following steps: identifying the actual annotation box of the point cloud data; expanding the actual annotation box to generate a simulated annotation box; based on the point cloud data, obtaining the points that fall into the simulated annotation box; calculating the number of points that fall into the simulated annotation box, and when the number of points is greater than a preset number threshold, measuring the degree of fit between the point cloud data and the azimuth plane of the actual annotation box.

[0056] The technical principle of the present invention is to generate an enlarged virtual truth value on the basis of the annotation truth value, calculate according to the points within this virtual truth value, and compare with the pre-set threshold annotation to determine whether the annotation truth value meets the fit. The present invention can automatically judge and screen the unfit annotation truth values within the threshold definition, improve the inspection efficiency, and at the same time reduce the labor cost of the inspection.

[0057] To this end, referring to Figure 1 , an embodiment of the present invention provides a schematic flowchart of a data processing method for the edge attachment of a point cloud annotation box in autonomous driving. This method can be executed by a data processing device for the edge attachment of a point cloud annotation box in autonomous driving. The device can be implemented in the form of hardware and / or software, and the device can be configured in a computer. As Figure 1 shown, this method includes:

[0058] S100: Identify the actual annotation box of the point cloud data.

[0059] It can be understood that in the autonomous driving scenario, the point cloud data is usually collected by a lidar (LiDAR) to form a set of points in a three-dimensional space. The actual annotation box refers to the bounding box of the target object manually or semi-automatically annotated by the annotator according to the point cloud data, and usually includes the position, size and orientation information of the object. The goal of this step is to extract the information of these annotation boxes from the point cloud data to provide a basis for subsequent processing.

[0060] In the autonomous driving scenario in the art, the point cloud data collected by the lidar usually contains multiple target objects, and each object is surrounded by an annotation box. The identification of the actual annotation box can be achieved through the following methods:

[0061] 1. Data preprocessing: Extract point cloud information from lidar data, which usually includes the three-dimensional coordinates (x, y, z) of each point and other attributes (such as intensity, timestamp, etc.).

[0062] 2. Annotation box extraction: Use existing annotation information (such as object category, bounding box coordinates, etc.) to extract the actual annotation box of each target object from the point cloud data. The annotation box usually exists in the form of the minimum bounding rectangle and can accurately enclose the target object.

[0063] 3. Multimodal data fusion: In practical applications, combining multimodal data such as cameras and radars can further improve the accuracy of annotation box recognition.

[0064] S200: Expand the actual annotation box to generate a simulated annotation box.

[0065] It can be understood that, in order to more comprehensively evaluate the fitting degree between the annotation box and the point cloud, the present invention innovatively expands the actual annotation box to generate a larger simulated annotation box. The expansion range can be adjusted according to the type of object and annotation requirements, such as increasing a certain proportion of the size or a fixed distance. The role of the simulated annotation box is to contain more point cloud data that may be related to the target object, so as to achieve a more accurate automatic evaluation of the rationality of the annotation box.

[0066] In a specific implementation, the azimuth planes of the actual annotation box include the front, back, left, right, top, and bottom. In the data processing method for the point cloud annotation box to be adjacent to the edge in autonomous driving, the azimuth planes of the actual annotation box include the front, back, left, right, top, and bottom. These azimuth planes jointly define the geometric shape and position of the annotation box in three-dimensional space and are an important basis for evaluating the fitting degree between the point cloud data and the annotation box. The following is a detailed description of each azimuth plane and how to use these azimuth planes in the algorithm to calculate the fitting degree:

[0067] 1. Definition of azimuth plane

[0068] The actual annotation box is a three-dimensional bounding box (3D Bounding Box), which is usually defined by the following parameters:

[0069] Center point coordinates: (x, y, z)

[0070] Dimensions: Length (L), width (W), height (H)

[0071] Orientation: The rotation angle of the annotation box (usually based on a certain axis)

[0072] Based on these parameters, the six azimuth planes of the annotation box can be defined as:

[0073] Front: The front surface of the annotation box, perpendicular to the length direction of the annotation box.

[0074] Back: The rear surface of the annotation box, parallel and opposite to the front.

[0075] Left side: The left surface of the annotation box, perpendicular to the width direction of the annotation box.

[0076] Right side: The right surface of the annotation box, parallel and opposite to the left side.

[0077] Top surface: The upper surface of the annotation box, perpendicular to the height direction of the annotation box.

[0078] Bottom surface: The lower surface of the annotation box, parallel and opposite to the top surface.

[0079] Specifically, the present invention provides a specific example of expanding the original annotation box to generate a simulated annotation box, specifically as follows:

[0080] S210: Expand the front, back, left, right, and top surfaces of the actual annotation box outward by 30 cm respectively.

[0081] In the autonomous driving scenario, the point cloud annotation box is usually a three-dimensional rectangular box used to represent the spatial range of target objects (such as vehicles, pedestrians, etc.). To more comprehensively evaluate the fitting degree between the annotation box and the point cloud, it is necessary to expand the five surfaces (front, back, left, right, and top) of the annotation box outward.

[0082] S220: Shrink the bottom surface of the actual annotation box inward by 50 cm.

[0083] For the bottom surface of the annotation box, shrinking it inward by 50 cm is to avoid including ground point clouds unrelated to the target object in the evaluation range, thereby improving the accuracy of the fitting degree calculation.

[0084] S230: Generate a simulated annotation box.

[0085] S300: Based on the point cloud data, obtain the points that fall into the simulated annotation box.

[0086] By calculating the relationship between the coordinates of each point in the point cloud data and the simulated annotation box, all the points that fall into the simulated annotation box are screened out. This step can be achieved through simple geometric judgment, such as checking whether the coordinates of the points are within the boundary range of the simulated annotation box. These points will be used for the subsequent fitting degree calculation.

[0087] S400: Calculate the number of points that fall into the simulated annotation box. When the number of points is greater than the preset number threshold, measure the fitting degree between the point cloud data and the azimuth surface of the actual annotation box.

[0088] In the present invention, the number of points is calculated as follows: count the number of point clouds falling within the simulated annotation box. If the number of points is less than the preset threshold, the true state of the object cannot be recognized, and it is meaningless to perform the annotation truth bordering judgment. Therefore, the bordering judgment is not performed by default, and the process ends.

[0089] Degree of fit calculation: When the number of points exceeds the threshold, further evaluate the degree of fit between the point cloud data and the azimuth plane of the actual annotation box. The degree of fit can be quantified by calculating geometric features such as the distance distribution and angle deviation between the point cloud and the boundary of the annotation box. For example, use the distance statistics method from points to the plane, or the matching degree between the normal distribution of the point cloud and the normal of the annotation box.

[0090] In a specific implementation, the preset quantity threshold is 100. When the number of points is greater than 100, calculate the degree of fit between the point cloud data and the azimuth plane of the actual annotation box.

[0091] In a specific implementation, calculating the degree of fit between the point cloud data and the azimuth plane of the actual annotation box specifically includes:

[0092] S410: Calculate the vertical distance from the points within the simulated annotation box to each azimuth plane of the actual annotation box.

[0093] In a specific implementation, the point cloud data is usually collected by a Light Detection and Ranging (LiDAR). The annotation box is used to represent the spatial range of a target object (such as a vehicle, a pedestrian, etc.). First, it is necessary to calculate the vertical distance from each point within the simulated annotation box to the six azimuth planes (front, back, left, right, top, bottom) of the actual annotation box. This step can be achieved through geometric calculations. For example, for a plane equation Ax + By + Cz + D = 0, the vertical distance from the point (x 0 , y 0 , z 0 ) to the plane is: A, B, C, and D are the coefficients of the plane equation, representing the normal vector of the plane. The normal vector of the plane is a vector perpendicular to the plane, and its direction is determined by the geometric shape of the plane.

[0094] S420: Count the number of points whose distance range from the azimuth plane is within 30 cm.

[0095] For each azimuth plane, count the number of points whose distance is within 30 cm. The purpose of this step is to screen out the point cloud data close to the surface of the annotation box, providing a basis for the subsequent fit degree judgment.

[0096] S430: If the number of points with an outer distance within 30 cm is greater than 5, it is determined that the point cloud data does not fit this azimuth plane.

[0097] It is understandable that if the number of points within 30 cm from the outside of a certain azimuth plane is greater than 5, it indicates that there is a relatively large amount of point cloud data close to the annotation box outside this azimuth plane, but it is not covered by the annotation box. This may be because there is a deviation between the boundary of the annotation box and the boundary of the actual object, resulting in some point clouds not being included in the annotation box. Therefore, it is determined that the point cloud data does not fit this azimuth plane.

[0098] S440: If the number of points with an inner distance of 30 cm is less than 5, it is determined that the point cloud data does not fit this azimuth plane.

[0099] It is understandable that if the number of points within 30 cm from the inside of a certain azimuth plane is less than 5, it indicates that the degree of fit between the point cloud data and the annotation box is insufficient inside this azimuth plane. This may be because there is a deviation between the boundary of the annotation box and the boundary of the actual object, resulting in some point clouds being wrongly included in the annotation box. Therefore, it is determined that the point cloud data does not fit this azimuth plane.

[0100] S450: If the number of points with an outer distance of 30 cm is less than or equal to 5, and the number of points with an inner distance of 30 cm is greater than or equal to 5, it is determined that the point cloud data fits this azimuth plane.

[0101] It is understandable that if the number of points within 30 cm from the outside of a certain azimuth plane is less than or equal to 5, and the number of points within 30 cm from the inside of this azimuth plane is greater than or equal to 5, it means that: the number of points within 30 cm from the outside is small, indicating that the boundary of the annotation box is basically consistent with the boundary of the actual object, and not too much point cloud data is missed. The number of points within 30 cm from the inside is large, indicating that the boundary of the annotation box fits well with the point cloud data, and not too many point clouds are wrongly included. Therefore, considering the above situations, it is determined that the point cloud data fits well with this azimuth plane.

[0102] It is understandable that this solution is to solve the fitting problem of the point cloud annotation true three-dimensional box. There are mainly three key technologies: 1 is to perform an outward expansion on the front, back, left, right, and top surfaces of the three-dimensional box, and an inward contraction on the bottom surface to form a virtual annotation three-dimensional box. 2 is to make a judgment on the edge fitting calculation, and determine the fitting degree by the number of points at a fixed distance on both the inside and outside of the actual annotation surface. 3 is to screen the fitting surfaces, and select the azimuth planes that need to be fitted based on the position of the annotation box relative to the acquisition vehicle.

[0103] Furthermore, the present invention provides the following embodiments:

[0104] S1. Based on the actual annotation true value data, expand the front, back, left, right, and top 5 surfaces of the annotation true three-dimensional box by 30 cm, and contract the bottom surface of the annotation true three-dimensional box inward by 50 cm to form a virtual annotation true value.

[0105] S2. Recalculate the point cloud data to obtain the point cloud that falls into the virtual annotation ground truth box.

[0106] S3. Calculate the number of point clouds in the virtual annotation ground truth box. If the number of point clouds is less than or equal to 100 and the true state of the object cannot be recognized, it is meaningless to perform the ground truth edge attachment judgment, so the edge attachment judgment is not performed by default. If the number of point clouds is greater than 100, then perform the edge attachment judgment on the annotation ground truth.

[0107] S4. During the lidar scanning process, occlusion may occur, resulting in sparse or even no side point clouds of some orientations of the entity object. If the edge attachment judgment is performed on these orientations, there will be many misjudgments. Therefore, it is necessary to calculate the orientation plane for which the edge attachment judgment needs to be performed according to the position relationship between the point cloud and the vehicle itself.

[0108] The relative vehicle orientation takes the vehicle's due front as the x-axis, the left side as the y-axis, and the angle of rotation around the x-axis to the y-axis is the yaw angle. The vehicle has 5 faces: front, rear, left, right, and top.

[0109] yaw angle detection surface (-22.5,22.5] rear, top (22.5,67.5] rear, left, top (67.5,112.5] left, top (112.5,157.5] front, left, top (157.5, 180] and (-180, -157.5) front, top (-157.5,-112.5] front, right, top (-112.5,-67.5] right, top (-67.5,-22.5] right, rear, top

[0110] S5. For the orientation plane for which the edge attachment judgment needs to be performed, calculate the perpendicular distance from the points in the virtual annotation ground truth box to the orientation plane, and count the data volume of the points within a distance range of 30 cm.

[0111] S6. If the number of points with an outer distance within 30 cm is greater than 5, it is judged that the orientation plane does not fit. If the number of points with an inner distance within 30 cm is less than 5, it is judged that the orientation plane does not fit. If the number of points with an outer distance within 30 cm is less than or equal to 5 and the number of points with an inner distance within 30 cm is greater than or equal to 5, it is judged that the orientation plane fits.

[0112] S7. Present the data of the annotation ground truth that does not fit in the form of a list for the quality inspection personnel to process.

[0113] The algorithm strategy solution for the edge attachment of the point cloud annotation box in the autonomous driving described in the present invention is applied to the fitting inspection of the point cloud annotation ground truth three-dimensional box. Compared with the fitting inspection done manually, through this automated ground truth fitting inspection, the inspection efficiency is greatly improved, and at the same time, the labor cost for inspection is reduced.

[0114] As Figure 3 shown, the present invention also provides a data processing device for the edge attachment of the point cloud annotation box in the autonomous driving. The data processing device includes:

[0115] An identification module, which is used to identify the actual annotation box of the point cloud data;

[0116] A simulation module, which is used to expand the actual annotation box to generate a simulated annotation box;

[0117] An acquisition module, which is used to acquire points falling into the simulated annotation box based on the point cloud data;

[0118] A calculation module, which is used to calculate the number of points falling into the simulated annotation box. When the number of points is greater than a preset number threshold, measure the degree of fit between the point cloud data and the azimuth plane of the actual annotation box.

[0119] Preferably, the simulation module expands the original annotation box to generate a simulated annotation box, specifically:

[0120] The azimuth planes of the actual annotation box include the front, back, left, right, top, and bottom;

[0121] Expand the front, back, left, right, and top of the actual annotation box outward by 30 cm respectively;

[0122] Shrink the bottom of the actual annotation box inward by 50 cm;

[0123] Generate a simulated annotation box.

[0124] Preferably, the preset number threshold is 100. When the number of points is greater than 100, measure the degree of fit between the point cloud data and the azimuth plane of the actual annotation box.

[0125] Figure 4 FIG. shows a schematic structural diagram of an electronic device 10 that can be used to implement an embodiment of the present invention. The electronic device is intended to represent various forms of digital computers, such as, laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as, personal digital processors, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present invention described herein and / or claimed.

[0126] As Figure 4As shown, the electronic device 10 includes at least one processor 11 and a memory communicatively connected to the at least one processor 11, such as read-only memory (ROM) 12, random access memory (RAM) 13, etc. The memory stores a computer program executable by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the electronic device 10 can also be stored. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. The input / output (I / O) interface 15 is also connected to the bus 14.

[0127] Multiple components in the electronic device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disc, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0128] The processor 11 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the processor 11 include but are not limited to a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as solving a data processing method for the edge attachment of a point cloud annotation box in autonomous driving.

[0129] In some embodiments, a data processing method for the edge attachment of a point cloud annotation box in autonomous driving can be implemented as a computer program tangibly embodied in a computer-readable storage medium, such as the storage unit 18. In some embodiments, part or all of the computer program can be loaded and / or installed onto the electronic device 10 via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the data processing method for the edge attachment of a point cloud annotation box in autonomous driving described above can be executed. Alternatively, in other embodiments, the processor 11 can be configured to execute a data processing method for the edge attachment of a point cloud annotation box in autonomous driving in any other appropriate manner (e.g., by means of firmware).

[0130] The various embodiments of the systems and techniques described above in this specification can be implemented in digital electronic circuitry, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on a chip (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include: being implemented in one or more computer programs that are executable and / or interpretable on a programmable system including at least one programmable processor, which can be a special-purpose or general-purpose programmable processor that receives data and instructions from, and transmits data and instructions to, a storage system, at least one input device, and at least one output device.

[0131] The computer programs for implementing the methods of the present invention can be written in any combination of one or more programming languages. These computer programs can be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus, such that the computer programs, when executed by the processor, cause the functions / operations specified in the flowchart and / or block diagram to be implemented. The computer programs can be executed entirely on the machine, partly on the machine, as a stand-alone software package partly on the machine and partly on a remote machine or entirely on the remote machine or server.

[0132] In the context of the present invention, a computer-readable storage medium can be a tangible medium that can contain or store a computer program for use by or in connection with an instruction execution system, apparatus, or device. The computer-readable storage medium can include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. Alternatively, the computer-readable storage medium can be a machine-readable signal medium. More specific examples of the machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0133] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and a pointing device (e.g., a mouse or a trackball), by which the user can provide input to the electronic device. Other kinds of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and the input received from the user can be in any form (including acoustic input, voice input, or tactile input).

[0134] The systems and techniques described herein can be implemented in a computing system including backend components (e.g., as a data server), or a computing system including middleware components (e.g., an application server), or a computing system including frontend components (e.g., a user computer having a graphical user interface or a web browser through which the user can interact with an implementation of the systems and techniques described herein), or a computing system including any combination of such backend components, middleware components, or frontend components. The components of the system can be interconnected to each other by digital data communication in any form or medium (e.g., a communication network). Examples of communication networks include: local area network (LAN), wide area network (WAN), blockchain network, and the Internet.

[0135] The computing system can include a client and a server. The client and the server are generally far from each other and usually interact through a communication network. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or a cloud host, which is a host product in the cloud computing service system, solving the defects of difficult management and weak business scalability existing in traditional physical hosts and VPS services.

[0136] An embodiment of the present invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements a data processing method for edge pasting of point cloud annotation boxes in an autonomous driving as provided by the embodiment of the present invention.

[0137] In the process of implementing the computer program product, computer program code for performing the operations of the present invention can be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as an independent software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any kind of network - including a local area network (LAN) or a wide area network (WAN) - or, alternatively, can be connected to an external computer (e.g., by using an Internet service provider to connect through the Internet).

[0138] It should be understood that the various forms of the processes shown above can be used, steps can be reordered, added or deleted. For example, the steps recited in the present invention can be executed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved, and no limitation is imposed herein.

[0139] The above specific embodiments do not constitute a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims

1. A data processing method for edge-fitting of point cloud annotation boxes in autonomous driving, characterized in that: The data processing method comprises the following steps: Identify the actual annotation box of the point cloud data; Expand the actual annotation frame to generate a simulated annotation frame; Based on the point cloud data, obtain the points that fall into the simulation annotation box; The number of points falling into the simulated annotation frame is calculated. When the number of points is greater than the preset threshold, the degree of fit between the point cloud data and the orientation plane of the actual annotation frame is measured.

2. The method for processing data of point cloud annotation frame border in autonomous driving according to claim 1, characterized in that: The original annotation box is expanded to generate a simulated annotation box, specifically: The orientation planes of the actual annotation frame include the front, back, left, right, top and bottom; The front, back, left, right and top of the actual marking frame are respectively expanded outward by 30 cm; The bottom surface of the actual marking frame is shrunk inward by 50 cm; Generate a simulated callout box.

3. The method for processing data of point cloud annotation frame border in automatic driving according to claim 1, characterized in that: The preset number threshold is 100. When the number of points is greater than 100, the degree of fit between the point cloud data and the orientation plane of the actual annotation box is calculated.

4. The method for processing data of point cloud annotation frame border in autonomous driving according to claim 3, characterized in that: The degree of fit between the measured point cloud data and the orientation plane of the actual annotation frame specifically includes: Calculate the vertical distance from the point in the simulated annotation box to each azimuth plane of the actual annotation box; Count the number of points within 30 cm of the azimuth plane; If the number of points with an outer distance of 30 cm is greater than 5, it is judged that the point cloud data does not fit the azimuth surface; If the number of points with an inner distance of 30 cm is less than 5, it is judged that the point cloud data does not fit the azimuth surface; If the number of points at an outer distance of 30 cm is less than or equal to 5, and the number of points at an inner distance of 30 cm is greater than or equal to 5, it is determined that the point cloud data is aligned with the orientation surface.

5. The method for processing data of point cloud annotation frame border in automatic driving according to claim 4, characterized in that: When the number of points is less than the preset threshold, the process ends.

6. A data processing device for edge-fitting of point cloud annotation boxes in autonomous driving, characterized in that: The data processing device is configured to execute the data processing method for the point cloud annotation box edge in the autonomous driving, and the data processing device includes: A recognition module, which is used to identify the actual annotation box of the point cloud data; A simulation module, which is used to expand the actual annotation box to generate a simulation annotation box; An acquisition module, which is used to acquire points falling into a simulation annotation box based on point cloud data; The calculation module is used to calculate the number of points falling into the simulated annotation frame. When the number of points is greater than a preset number threshold, the degree of fit between the point cloud data and the orientation plane of the actual annotation frame is measured.

7. The data processing device for edge-fitting of point cloud annotation frames in automatic driving according to claim 6, characterized in that: The simulation module expands the original annotation frame to generate a simulation annotation frame, specifically: The orientation planes of the actual annotation frame include the front, back, left, right, top and bottom; The front, back, left, right and top of the actual marking frame are respectively expanded outward by 30 cm; The bottom surface of the actual marking frame is shrunk inward by 50 cm; Generate a simulated callout box.

8. The data processing device for edge-fitting of point cloud annotation frames in automatic driving according to claim 6, characterized in that: The preset number threshold is 100. When the number of points is greater than 100, the degree of fit between the point cloud data and the orientation plane of the actual annotation box is calculated.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively coupled to the at least one processor; In which, the memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the data processing method for point cloud annotation box adhesion in autonomous driving as described in any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that: The computer readable storage medium stores a computer program, The computer program is used to enable a processor to implement a data processing method for edge-fitting of a point cloud annotation box in autonomous driving as described in any one of claims 1 to 5 when executed.

Images