Real-time denoising method and device for laser point cloud data, electronic equipment and storage medium
By performing coarse and fine filtering on the laser point cloud data, noise in the lidar is removed in adverse weather conditions, solving the problem of decreased lidar detection accuracy and achieving high-precision perception under conditions such as heavy fog.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- ZHIDAO NETWORK TECH (BEIJING) CO LTD
- Filing Date
- 2023-04-24
- Publication Date
- 2026-06-19
AI Technical Summary
In adverse weather conditions such as heavy fog, dense smoke, and heavy rain, the detection accuracy of lidar decreases, affecting the perception capabilities of autonomous vehicles.
The laser point cloud data is denoised using coarse and fine filtering methods. First, coarse filtering removes most of the noise, and then fine filtering removes the remaining noise. The quadratic function relationship between reflection intensity and distance is used to distinguish noise from the target. Raster processing is used to determine the noise area for fine filtering.
Maintaining the detection accuracy of lidar under adverse weather conditions reduces the probability of false target detection, ensuring the stability and real-time performance of vehicle environmental perception.
Smart Images
Figure CN116485674B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of autonomous driving technology, and in particular to a real-time denoising method, apparatus, electronic device, and storage medium for laser point cloud data. Background Technology
[0002] LiDAR, also known as optical radar, is short for laser detection and ranging system. It detects the distance and position of surrounding objects by emitting and receiving laser beams, thereby enabling autonomous vehicles to perceive their surrounding environment.
[0003] LiDAR boasts advantages such as high measurement accuracy, fast response speed, and strong anti-interference capability, providing high-precision target position and contour information, making it highly effective for tasks like target classification and recognition. However, in adverse weather conditions such as heavy rain, dense smoke, and heavy fog, the detection accuracy of LiDAR drops rapidly, impacting the perception results of autonomous vehicles. Summary of the Invention
[0004] This application provides a real-time denoising method, apparatus, electronic device, and storage medium for laser point cloud data, to denoise laser point cloud data in special scenarios and reduce the impact on perception results.
[0005] The embodiments of this application adopt the following technical solutions:
[0006] In a first aspect, embodiments of this application provide a real-time denoising method for laser point cloud data, wherein the method includes:
[0007] Acquire laser point cloud data generated by a target lidar, which is mounted on a vehicle;
[0008] The laser point cloud data is coarsely filtered to obtain the first denoising result;
[0009] The first denoising result is used for fine filtering to obtain a second denoising result, which is then used as the denoising result for the laser point cloud data.
[0010] In some embodiments, the method further includes:
[0011] The target lidar is pre-calibrated, and a working mode for a preset usage scenario is established, wherein the preset usage scenario includes at least one of the following: severe weather scenarios such as fog, snow, or rain.
[0012] In some embodiments, the laser point cloud data is coarsely filtered to obtain a first denoising result, including:
[0013] When the working mode of the preset usage scenario is turned on, the filtering range of the first target noise is set in the current usage scenario;
[0014] Based on the filtering range of the first target noise and the dynamic threshold of the noise reflection intensity, the first target noise is filtered out to obtain the first denoising result.
[0015] In some embodiments, fine filtering is performed based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data, including:
[0016] Based on the first denoising result, determine the region where the second target noise is located;
[0017] By using a preset height dynamic threshold, the second target noise in the region where the second target noise is located is filtered out to obtain the second denoising result.
[0018] In some embodiments, the method further includes:
[0019] The dynamic threshold of the noise reflection intensity is set as a quadratic function of the distance between the first target noise and the target lidar, and then fitted.
[0020] Based on the fitting results, points whose point cloud intensity values are less than the dynamic threshold of noise reflection intensity are removed as noise.
[0021] In some embodiments, the step of performing fine filtering based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data includes:
[0022] The point cloud data in the first denoising result is divided into multiple grids in the XOY plane according to the x range and y range;
[0023] Calculate the number of noise points in each grid cell;
[0024] If the number of noise points exceeds a preset ratio, then the region corresponding to the current grid is finely filtered according to the height sorting result of the noise points in the current grid.
[0025] In some embodiments, acquiring laser point cloud data generated by the target lidar includes:
[0026] Acquire the laser point cloud data generated by the rotating mechanical lidar in each frame;
[0027] The denoising results of the laser point cloud data include:
[0028] Point cloud noise in the laser point cloud data generated in each frame of the rotating mechanical lidar is filtered out.
[0029] Secondly, embodiments of this application also provide a point cloud detection method in a foggy scene, wherein the point cloud detection method includes:
[0030] The noise removal method for laser point cloud data described in the first aspect is used to filter out noise in the foggy scene;
[0031] Based on the noise filtering results in the aforementioned foggy scene, the perception results of the target point cloud data are obtained.
[0032] Thirdly, embodiments of this application also provide a real-time denoising device for laser point cloud data, wherein the device includes:
[0033] An acquisition module is used to acquire laser point cloud data generated by a target lidar, which is installed on a vehicle.
[0034] The first denoising module is used to perform coarse filtering on the laser point cloud data to obtain a first denoising result;
[0035] The second denoising module is used to perform fine filtering based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data.
[0036] Fourthly, embodiments of this application also provide an electronic device, including: a processor; and a memory arranged to store computer-executable instructions, which, when executed, cause the processor to perform the above-described method.
[0037] Fifthly, embodiments of this application also provide a computer-readable storage medium that stores one or more programs, which, when executed by an electronic device including multiple applications, cause the electronic device to perform the above-described method.
[0038] The at least one technical solution adopted in this application embodiment can achieve the following beneficial effects: For special scenarios (heavy fog weather), firstly, the laser point cloud data generated by the target lidar is acquired, then the laser point cloud data is coarsely filtered to obtain a first denoising result, and then finely filtered based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data. Since each frame of laser point cloud data acquired by the target lidar installed on the vehicle undergoes real-time double denoising before being used for obstacle and other target perception, the stability of the vehicle's environmental perception results is ensured.
[0039] In addition, the removal of noise also reduces the probability of false detection of targets. Attached Figure Description
[0040] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:
[0041] Figure 1This is a schematic diagram of the real-time noise reduction method for laser point cloud data in the embodiments of this application;
[0042] Figure 2 This is a schematic diagram of the distribution of point cloud data in a noisy scene in an embodiment of this application;
[0043] Figure 3 This is a diagram showing the effect after coarse filtering in the embodiments of this application;
[0044] Figure 4 This is a magnified schematic diagram of residual noise in an embodiment of this application (indicated by arrows);
[0045] Figure 5 This is a diagram showing the effect after fine filtration in the embodiments of this application;
[0046] Figure 6 This is a schematic diagram of the real-time noise reduction device for laser point cloud data in an embodiment of this application;
[0047] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation
[0048] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0049] A lidar mainly consists of a laser emitter, a receiver, a scanner, a lens antenna, and a signal processing circuit. According to the scanning method, lidar can be divided into MEMS lidar, Flash lidar, phased array lidar, and mechanical rotating lidar. To ensure the applicability of the solution, the acquisition of laser point cloud data in this application embodiment adopts mechanical rotating lidar.
[0050] Furthermore, LiDAR can provide high-resolution maps, enabling vehicles to achieve high-precision positioning and path planning. LiDAR also has a wide field of view, allowing it to detect targets from multiple angles. In complex environments, LiDAR can provide more accurate target location information, enabling autonomous vehicles to avoid collisions.
[0051] The technical solutions provided by the various embodiments of this application are described in detail below with reference to the accompanying drawings.
[0052] This application provides a real-time denoising method for laser point cloud data, such as... Figure 1 The diagram illustrates a real-time denoising method for laser point cloud data in an embodiment of this application. The method includes at least the following steps S110 to S140:
[0053] Step S110: Acquire laser point cloud data generated by the target lidar, which is installed on the vehicle.
[0054] The target lidar comprises multiple units, each deployed on the vehicle. LiDAR is a comprehensive optical detection and measurement system that emits and receives laser beams, analyzes the return time of the laser after encountering a target object, and calculates the relative distance between the target object and the vehicle. Currently, 16-line, 32-line, and 64-line lidars are common. The more lidar beams, the higher the measurement accuracy and the higher the security. Taking a mechanically rotating lidar as an example, the laser point cloud data generated by the target lidar is acquired in real time. Furthermore, since mechanically rotating lidar uses surround-view perception, it can be approximated as isotropic in the direction of each beam, meaning that, without considering positive and negative directions, the imaging distance of the laser point cloud data obtained by the lidar scan is the same in the same direction.
[0055] Step S120: Perform coarse filtering on the laser point cloud data to obtain the first denoising result.
[0056] The coarse filtering of the laser point cloud data primarily removes most of the noisy point clouds within a preset range. During this coarse filtering, since the reflection intensity of nearby targets is significantly higher than that of noise points in a preset scene (such as foggy weather), filtering can be based on intensity. In other words, the reflection intensity in the preset scene is low, and a threshold can be set for filtering.
[0057] It is understood that the target lidar may have the ability to filter out noise in a preset scene, or may be turned on or off according to the preset scene.
[0058] The first denoising result is the processed result after initial filtering, where noise has been removed. At this point, a small amount of noise still remains in the first denoising result. For these small noise points, the areas or networks with more noise can be identified by traversing the grid, thus enabling fine filtering.
[0059] It's important to note that the challenge in this step lies in distinguishing between the target and noise. Since the reflected signal intensity of nearby targets is significantly higher than that of noise in a foggy scene, filtering can be based on signal intensity. Experimental research has shown that the reflected signal intensity and the energy received by the lidar have a quadratic function relationship, and the intensity decreases with distance. This allows us to differentiate between noise and the actual target, thus identifying the corresponding target or noise. Typically, the reflected signal intensity of the target is greater than that of noise; therefore, the energy received by the lidar can be used to distinguish between noise and the target.
[0060] Step S130: Perform fine filtering based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data.
[0061] Based on the first denoising result obtained in the previous step, further fine filtering is performed. After fine filtering, a second denoising result is obtained. At this point, there are no residual noise points in the second denoising result, and it can be used as the denoising result of the laser point cloud data.
[0062] It is important to note that the difficulty in this step lies in determining which region of the point cloud data contains the most noise. In practice, this is achieved by dividing the laser point cloud data of the current frame into grids of equal size, and then traversing each grid to identify the regions containing the grids that need to be filtered out again.
[0063] It's understandable that the same processing method is applied to each frame of laser point cloud data from the LiDAR. Through the above steps, the processing speed for each frame can be reduced to milliseconds, thus effectively meeting real-time requirements.
[0064] By employing the aforementioned dual-stage filtering method to remove laser point cloud noise, the lidar can maintain good detection accuracy even in adverse weather conditions such as heavy rain, dense smoke, and heavy fog. Furthermore, any noise remaining from the coarse filtering stage can be accurately removed during the fine filtering stage.
[0065] Unlike related technologies where the perception accuracy of lidar is affected by scenarios such as heavy fog, the method described above uses coarse and fine filtering to remove noise from each frame of lidar data. Furthermore, the response time is at the millisecond level, thus meeting the requirements for real-time performance.
[0066] Using the above methods, noise removal mode can be actively activated on the LiDAR in scenarios such as heavy fog, thereby improving or maintaining the perception accuracy of the LiDAR in such scenarios.
[0067] In one embodiment of this application, the method further includes: pre-calibrating the target lidar and establishing a working mode for a preset usage scenario, wherein the preset usage scenario includes at least one of the following: severe weather scenarios such as fog, snow, or rain.
[0068] The target LiDAR installed on the vehicle needs to be pre-calibrated, and then a working mode for a preset usage scenario needs to be established. It can be understood that the working mode for these preset usage scenarios can be implemented through software programs and adapted according to the different types of target LiDARs.
[0069] In one embodiment of this application, coarse filtering of the laser point cloud data to obtain a first denoising result includes: when the working mode of the preset usage scenario is turned on, setting the filtering range of the first target noise in the current usage scenario; and filtering out the first target noise according to the filtering range of the first target noise and the dynamic threshold of the noise reflection intensity to obtain the first denoising result.
[0070] After one frame of scanning, the laser point cloud data acquired by the lidar is as follows: Figure 2 As shown. After activating the preset usage scenario working mode, set the filtering range for the first target noise in the current usage scenario. For example, for setting the noise range in foggy weather, based on experience, the range is generally within 0-10m, so the filtering range can be set to 0-15m. This ensures that all possible noise points are covered during the initial filtering.
[0071] Furthermore, based on the relationship between the filtering range of the first target noise and the dynamic threshold of the noise reflection intensity, the first target noise is filtered out to obtain the first denoising result. In this way, most of the noise can be filtered out.
[0072] For example, during the coarse filtering process, most of the noise is filtered out based on a dynamic threshold of reflection intensity. Because the reflection intensity of nearby targets is significantly higher than that of noise in a foggy scene, filtering can be based on intensity.
[0073] Experimental research revealed the following relationship between the intensity of reflected signals in lidar and distance:
[0074]
[0075] Where ρ is a coefficient, Ti is the transmittance of various materials, r is the distance, and Q_received is the received energy. Since the relationship between the point cloud intensity and the received energy varies among different LiDAR service providers, they can be considered to be linear. Therefore, the dynamic threshold of noise is designed as a quadratic function of distance for fitting, and points less than the threshold are considered noise and deleted.
[0076] For the quadratic function y = ax 2 +bx+c, the received energy and distance are related as a quadratic function. Since the mechanically rotating lidar is approximately isotropic, all odd-degree terms are set to 0, thus we get y = a·x 2 +c, where a and c are coefficients (empirical values obtained based on different lidar signals), x is the distance from the lidar, and y is the threshold of reflection intensity.
[0077] The point cloud data collected by the mechanical rotating lidar has a symmetrical relationship. When the mechanical rotating lidar is installed on a vehicle, there is also a corresponding distance relationship for each opposite direction of the vehicle's coordinate system in the front, back, left, and right views. That is, the reflection intensity is only related to the distance and is not related to the sign. Therefore, the odd-numbered terms are zero and the squared terms are not zero.
[0078] After coarse filtering, the black dots are identified as noise. For example... Figure 3 As shown.
[0079] In one embodiment of this application, a second denoising result is obtained by fine filtering based on a first denoising result and used as the denoising result of the laser point cloud data. This includes: determining the region where the second target noise is located based on the first denoising result; and filtering out the second target noise in the region where the second target noise is located by using a preset height dynamic threshold to obtain the second denoising result.
[0080] like Figure 4 The image shows the residual noise in the first denoising result. The arrow points to the area indicated by the noise.
[0081] The region where the second target noise is located is determined in the first denoising result, and different noise points can be distinguished by a preset height dynamic threshold.
[0082] It is understandable that the laser point cloud imaging result is a point cloud in a two-dimensional plane, which is a point cloud BEV (Bird's Eye View) view. The point cloud BEV view refers to the projection of the point cloud onto a plane perpendicular to the height direction.
[0083] In one embodiment of this application, the method further includes: setting the dynamic threshold of the noise reflection intensity as a quadratic function of the distance between the first target noise and the target lidar and fitting it; based on the fitting result, deleting points whose point cloud intensity values are less than the dynamic threshold of the noise reflection intensity as noise.
[0084] Based on the above, y = a·x 2+c, where a and c are coefficients, x is the distance to the radar, and y is the threshold for reflection intensity. The resulting quadratic function can be fitted, and then points with point cloud intensity values less than the dynamic threshold for noise reflection intensity can be removed as noise based on the fitting result.
[0085] By using the above method and the fitting results of the quadratic function, the relationship between point clouds of different intensities and the distance threshold range can be determined and filtered out as noise.
[0086] In one embodiment of this application, the step of fine filtering based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data includes: dividing the point cloud data in the first denoising result into multiple grids in the XOY plane according to the x range and y range; calculating the number of noise points in each grid; if the number of noise points exceeds a preset ratio value, then fine filtering is performed on the region corresponding to the current grid according to the height sorting result of the noise points in the current grid.
[0087] In areas with high noise levels, a high dynamic threshold is set to filter out residual noise points, which are sparse, high-reflectivity points. If the noise level is low, fine filtering is unnecessary; areas with high noise levels will have residual noise.
[0088] For example, the x-range and y-range are divided into grids. For each grid, the number of noise points within the grid is calculated. If the percentage exceeds a certain proportion, it indicates that the grid contains a lot of noise and requires further processing. The processing method is to sort the points within the grid by height and delete points whose height is below a certain percentage. For example, if the top 10% of noise points are filtered out, there are still many target points remaining. After fine filtering, the result is as follows: Figure 5 The diagram shows the effect.
[0089] In one embodiment of this application, acquiring laser point cloud data generated by a target lidar includes: acquiring the laser point cloud data generated by a rotating mechanical lidar for each frame; the denoising result of the laser point cloud data includes: filtering out point cloud noise from the laser point cloud data generated by the rotating mechanical lidar for each frame.
[0090] For rotating mechanical lidar, the laser point cloud data generated in each frame is acquired, and then point cloud noise is filtered out in real time. Since the computational cost per frame is on the order of milliseconds and the computational accuracy is high, the filtered result can be obtained in real time.
[0091] In embodiments of this application, a point cloud detection method in a foggy scene is also provided, wherein the point cloud detection method includes:
[0092] The aforementioned denoising method for laser point cloud data is used to filter out noise in foggy scenes;
[0093] Based on the noise filtering results in the aforementioned foggy scene, the perception results of the target point cloud data are obtained.
[0094] The real-time noise reduction method for the laser point cloud data includes:
[0095] Acquire laser point cloud data generated by a target lidar, which is mounted on a vehicle;
[0096] The laser point cloud data is coarsely filtered to obtain the first denoising result;
[0097] The first denoising result is used for fine filtering to obtain a second denoising result, which is then used as the denoising result for the laser point cloud data.
[0098] In specific implementation, the method further includes:
[0099] The target lidar is pre-calibrated, and a working mode for a preset usage scenario is established, wherein the preset usage scenario includes at least one of the following: severe weather scenarios such as fog, snow, or rain.
[0100] The laser point cloud data is coarsely filtered to obtain a first denoising result, including:
[0101] When the working mode of the preset usage scenario is turned on, the filtering range of the first target noise is set in the current usage scenario;
[0102] Based on the filtering range of the first target noise and the dynamic threshold of the noise reflection intensity, the first target noise is filtered out to obtain the first denoising result.
[0103] Based on the first denoising result, a second denoising result is obtained through fine filtering to serve as the denoising result for the laser point cloud data, including:
[0104] Based on the first denoising result, determine the region where the second target noise is located;
[0105] By using a preset height dynamic threshold, the second target noise in the region where the second target noise is located is filtered out to obtain the second denoising result.
[0106] The method further includes:
[0107] The dynamic threshold of the noise reflection intensity is set as a quadratic function of the distance between the first target noise and the target lidar, and then fitted.
[0108] Based on the fitting results, points whose point cloud intensity values are less than the dynamic threshold of noise reflection intensity are removed as noise.
[0109] The step of refining the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data includes:
[0110] The point cloud data in the first denoising result is divided into multiple grids in the XOY plane according to the x range and y range;
[0111] Calculate the number of noise points in each grid cell;
[0112] If the number of noise points exceeds a preset ratio, then the region corresponding to the current grid is finely filtered according to the height sorting result of the noise points in the current grid.
[0113] This application embodiment also provides a real-time noise reduction device 600 for laser point cloud data, such as... Figure 6 As shown, a schematic diagram of a real-time denoising device for laser point cloud data in an embodiment of this application is provided. The real-time denoising device 600 for laser point cloud data includes at least: an acquisition module 610, a first denoising module 620, and a second denoising module 630, wherein:
[0114] In one embodiment of this application, the acquisition module 610 is specifically used to: acquire laser point cloud data generated by a target lidar, wherein the target lidar is installed on a vehicle.
[0115] The target lidar comprises multiple units, each deployed on the vehicle. LiDAR is a comprehensive optical detection and measurement system that emits and receives laser beams, analyzes the return time of the laser after encountering a target object, and calculates the relative distance between the target object and the vehicle. Currently, 16-line, 32-line, and 64-line lidars are common. The more lidar beams, the higher the measurement accuracy and the higher the security. Taking a mechanically rotating lidar as an example, the laser point cloud data generated by the target lidar is acquired in real time. Furthermore, since mechanically rotating lidar uses surround-view perception, it can be approximated as isotropic in the direction of each beam, meaning that, without considering positive and negative directions, the imaging distance of the laser point cloud data obtained by the lidar scan is the same in the same direction.
[0116] In one embodiment of this application, the first denoising module 620 is specifically used to: perform coarse filtering on the laser point cloud data to obtain a first denoising result.
[0117] The coarse filtering of the laser point cloud data primarily removes most of the noisy point clouds within a preset range. During this coarse filtering, since the reflection intensity of nearby targets is significantly higher than that of noise points in a preset scene (such as foggy weather), filtering can be based on intensity. In other words, the reflection intensity in the preset scene is low, and a threshold can be set for filtering.
[0118] It is understood that the target lidar may have the ability to filter out noise in a preset scene, or may be turned on or off according to the preset scene.
[0119] The first denoising result is the processed result after initial filtering, where noise has been removed. At this point, a small amount of noise still remains in the first denoising result. For these small noise points, the areas or networks with more noise can be identified by traversing the grid, thus enabling fine filtering.
[0120] It is important to note that the key challenge in this step lies in distinguishing between the target and noise. Experimental research has shown that the intensity of the reflected signal and the energy received by the lidar have a quadratic relationship, and the signal attenuates with increasing distance. This allows for the differentiation between noise and the actual target, thus identifying the corresponding target or noise.
[0121] In one embodiment of this application, the second denoising module 630 is specifically used to: perform fine filtering based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data.
[0122] Based on the first denoising result obtained in the previous step, further fine filtering is performed. After fine filtering, a second denoising result is obtained. At this point, there are no residual noise points in the second denoising result, and it can be used as the denoising result of the laser point cloud data.
[0123] It is important to note that the difficulty in this step lies in determining which region of the point cloud data contains the most noise. In practice, this is achieved by dividing the laser point cloud data of the current frame into grids of equal size, and then traversing each grid to identify the regions containing the grids that need to be filtered out again.
[0124] It's understandable that the same processing method is applied to each frame of laser point cloud data from the LiDAR. Through the above steps, the processing speed for each frame can be reduced to milliseconds, thus effectively meeting real-time requirements.
[0125] It is understood that the above-mentioned real-time denoising device for laser point cloud data can realize each step of the real-time denoising method for laser point cloud data provided in the foregoing embodiments. The relevant explanations of the real-time denoising method for laser point cloud data are applicable to the real-time denoising device for laser point cloud data, and will not be repeated here.
[0126] Figure 7 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Please refer to it. Figure 7At the hardware level, the electronic device includes a processor, and optionally also includes an internal bus, a network interface, and memory. The memory may include main memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for other business operations.
[0127] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.
[0128] Memory is used to store programs. Specifically, programs may include program code, which includes computer operation instructions. Memory may include main memory and non-volatile memory, and provides instructions and data to the processor.
[0129] The processor reads the corresponding computer program from non-volatile memory into main memory and then runs it, forming a real-time denoising device for laser point cloud data at the logical level. The processor executes the program stored in memory and specifically performs the following operations:
[0130] Acquire laser point cloud data generated by a target lidar, which is mounted on a vehicle;
[0131] The laser point cloud data is coarsely filtered to obtain the first denoising result;
[0132] The first denoising result is used for fine filtering to obtain a second denoising result, which is then used as the denoising result for the laser point cloud data.
[0133] The above is as stated in this application. Figure 1The method executed by the real-time denoising device for laser point cloud data disclosed in the illustrated embodiment can be applied to a processor or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by integrated logic circuits in the processor's hardware or by instructions in software form. The processor can be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. It can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in the embodiments of this application can be directly manifested as execution by a hardware decoding processor, or execution by a combination of hardware and software modules in the decoding processor. The software module can reside in a mature storage medium in the field, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, or registers. This storage medium is located in memory, and the processor reads information from the memory and, in conjunction with its hardware, completes the steps of the above method.
[0134] The electronic device can also perform Figure 1 A method for implementing a real-time denoising device for laser point cloud data is described, and the real-time denoising device for laser point cloud data is implemented. Figure 1 The functions of the embodiments shown are not described in detail here.
[0135] This application also proposes a computer-readable storage medium that stores one or more programs, the programs including instructions that, when executed by an electronic device including multiple applications, enable the electronic device to perform... Figure 1 The method executed by the real-time denoising device for laser point cloud data in the illustrated embodiment is specifically used to perform:
[0136] Acquire laser point cloud data generated by a target lidar, which is mounted on a vehicle;
[0137] The laser point cloud data is coarsely filtered to obtain the first denoising result;
[0138] The first denoising result is used for fine filtering to obtain a second denoising result, which is then used as the denoising result for the laser point cloud data.
[0139] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0140] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. 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 illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.
[0141] 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.
[0142] 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.
[0143] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.
[0144] Memory may include non-persistent storage in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0145] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.
[0146] It should also be noted that 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.
[0147] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0148] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principle of this application should be included within the scope of the claims of this application.
Claims
1. A real-time denoising method for laser point cloud data, wherein, The method includes: Acquire laser point cloud data generated by a target lidar, which is mounted on a vehicle; The laser point cloud data is coarsely filtered to obtain the first denoising result; The laser point cloud data is coarsely filtered to obtain a first denoising result, including: Once the preset usage scenario working mode is enabled, the filtering range of the first target noise is set in the current usage scenario; Based on the filtering range of the first target noise and the dynamic threshold of the noise reflection intensity, the first target noise is filtered out to obtain the first denoising result; The dynamic threshold of the noise reflection intensity is set as a quadratic function of the distance between the first target noise and the target lidar, and then fitted. The quadratic function y = ax 2 +bx+c is used to characterize the quadratic function relationship between received energy and distance. Since the mechanically rotating lidar is approximately isotropic, the coefficients of all odd-degree terms are set to 0, thus obtaining the corrected quadratic function y = a·x. 2 +c, where a and c are empirical values obtained based on different lidar signals, x is the distance from the lidar, and y is the threshold of reflection intensity; the design basis of the modified quadratic function is the isotropic characteristics of the rotating mechanical lidar and the quadratic attenuation law of laser received energy with distance. Based on the fitting results, points whose point cloud intensity values are less than the dynamic threshold of noise reflection intensity are removed as noise. Based on the first denoising result, a second denoising result is obtained and used as the denoising result of the laser point cloud data; Based on the first denoising result, a second denoising result is obtained through fine filtering to serve as the denoising result for the laser point cloud data, including: The point cloud data in the first denoising result is divided into multiple grids in the XOY plane according to the x range and y range; Calculate the number of noise points in each grid cell; If the number of noise points exceeds a preset ratio, then the region corresponding to the current grid is finely filtered according to the height sorting result of the noise points in the current grid.
2. The method as described in claim 1, wherein, Based on the first denoising result, a second denoising result is obtained through fine filtering to serve as the denoising result for the laser point cloud data, including: Based on the first denoising result, determine the region where the second target noise is located; By using a preset height dynamic threshold, the second target noise in the region where the second target noise is located is filtered out to obtain the second denoising result.
3. The method as described in claim 1, wherein: Acquire laser point cloud data generated by the target lidar, including: Acquire the laser point cloud data generated by the rotating mechanical lidar in each frame; The denoising results of the laser point cloud data include: Point cloud noise in the laser point cloud data generated in each frame of the rotating mechanical lidar is filtered out.
4. A point cloud detection method in a foggy scene, wherein, The point cloud detection method includes: The noise removal method for laser point cloud data as described in any one of claims 1 to 3 is used to filter out noise in foggy scenes. Based on the noise filtering results in the aforementioned foggy scene, the perception results of the target point cloud data are obtained.
5. A real-time noise reduction device for laser point cloud data, wherein, The device includes: An acquisition module is used to acquire laser point cloud data generated by a target lidar, which is installed on a vehicle. The first denoising module is used to perform coarse filtering on the laser point cloud data to obtain a first denoising result; The laser point cloud data is coarsely filtered to obtain a first denoising result, including: Once the preset usage scenario working mode is enabled, the filtering range of the first target noise is set in the current usage scenario; Based on the filtering range of the first target noise and the dynamic threshold of the noise reflection intensity, the first target noise is filtered out to obtain the first denoising result; The dynamic threshold of the noise reflection intensity is set as a quadratic function of the distance between the first target noise and the target lidar, and fitted; the quadratic function y = ax 2 +bx+c is used to characterize the quadratic function relationship between received energy and distance. Since the mechanically rotating lidar is approximately isotropic, the coefficients of all odd-degree terms are set to 0, thus obtaining the corrected quadratic function y = a·x. 2 +c, where a and c are empirical values obtained based on different lidar signals, x is the distance from the lidar, and y is the threshold of reflection intensity; the design basis of the modified quadratic function is the isotropic characteristics of the rotating mechanical lidar and the quadratic attenuation law of laser received energy with distance. Based on the fitting results, points whose point cloud intensity values are less than the dynamic threshold of noise reflection intensity are removed as noise. The second denoising module is used to perform fine filtering based on the first denoising result to obtain a second denoising result as the denoising result of the laser point cloud data. Based on the first denoising result, a second denoising result is obtained through fine filtering to serve as the denoising result for the laser point cloud data, including: The point cloud data in the first denoising result is divided into multiple grids in the XOY plane according to the x range and y range; Calculate the number of noise points in each grid cell; If the number of noise points exceeds a preset ratio, then the region corresponding to the current grid is finely filtered according to the height sorting result of the noise points in the current grid.
6. An electronic device, comprising: processor; as well as A memory configured to store computer-executable instructions, which, when executed, cause the processor to perform the method of any one of claims 1 to 3.
7. A computer-readable storage medium storing one or more programs, which, when executed by an electronic device including a plurality of applications, cause the electronic device to perform the method of any one of claims 1 to 3.