Point cloud correction method, computer equipment and computer readable storage medium

Abstract

The embodiment of the invention discloses a point cloud correction method, computer equipment and a computer readable storage medium. The method comprises steps that a point cloud data set of forward and reverse scanning of a target plane by a laser radar is acquired, the point cloud data set comprises multiple frames of point clouds, a target gain coefficient is determined based on the point cloud data set, and the target gain coefficient is used for controlling distortion of the laser radar correction point clouds in a slow axis scanning direction; and controlling the laser radar to execute point cloud correction operation based on the target gain coefficient. According to the embodiment of the invention, the target gain coefficient can be determined based on the point cloud data set obtained through positive and negative scanning, so that the laser radar employs the target gain coefficient to correct the distortion of the point cloud in the slow-axis scanning direction, improves the layering condition of the point cloud during the positive and negative scanning of the laser, facilitates the improvement of the calculation precision of the point cloud, and improves the calculation efficiency. The problem of point cloud quality reduction caused by device feedback errors and the like is solved, and the precision and the stability of the point cloud are improved.

Description

Technical Field

[0001] This application relates to the field of point cloud correction technology, and in particular to a point cloud correction method, computer device, and computer-readable storage medium. Background Technology

[0002] The lidar is equipped with a galvanometer device. The lidar outputs slow-axis drive current and fast-axis drive current respectively, driving the galvanometer device to move around the slow axis or fast axis. When the direction of the slow-axis drive current switches, it can easily cause abnormal high-frequency signals. Abnormal high-frequency signals can cause distortion of the lidar's point cloud in the slow-axis scanning direction, which is not conducive to the lidar obtaining accurate and reliable point clouds. Summary of the Invention

[0003] One objective of this application is to provide a point cloud correction method, computer device, and computer-readable storage medium to improve the situation where point clouds in the slow-axis scanning direction are prone to distortion in related technologies.

[0004] In a first aspect, embodiments of this application provide a point cloud correction method applied to a lidar, the lidar including a galvanometer device, the point cloud correction method comprising: acquiring a point cloud dataset of the lidar scanning a target plane in both directions, the point cloud dataset including multiple frames of point clouds, each frame of the point cloud being obtained by the lidar using different gain coefficients to drive the galvanometer device to scan the target plane around the fast axis and slow axis respectively; determining a target gain coefficient based on the point cloud dataset, the target gain coefficient being used to control the lidar to correct distortions in the point cloud appearing in the slow axis scanning direction; and controlling the lidar to perform point cloud correction operations based on the target gain coefficient.

[0005] The embodiments of this application can determine the target gain coefficient based on the point cloud dataset obtained by forward and reverse scanning, so that the lidar can use the target gain coefficient to correct the distortion of the point cloud that is prone to occur in the slow axis scanning direction, improve the point cloud layering during forward and reverse laser scanning, thus improving the solution accuracy of the point cloud, solving the problem of point cloud quality reduction caused by device feedback errors, and improving the accuracy and stability of the point cloud.

[0006] In some embodiments, determining the target gain coefficient based on the point cloud dataset includes: determining the thickness of the point cloud with respect to the target plane; determining the minimum thickness among the thicknesses of the point cloud in multiple frames; and determining the gain coefficient corresponding to the minimum thickness as the target gain coefficient.

[0007] The minimum thickness reflects the minimum deviation between forward and reverse scanning. In this embodiment, the gain coefficient corresponding to the minimum thickness is used as the target gain coefficient. The lidar is controlled to correct the point cloud with low deviation and high accuracy according to the target gain coefficient, which helps to improve the solution accuracy of the point cloud.

[0008] In some embodiments, determining the thickness of the target plane based on the point cloud includes: extracting a forward scan point cloud set corresponding to the target plane from the point cloud; extracting a reverse scan point cloud set corresponding to the target plane from the point cloud; and determining the thickness of the target plane based on the forward scan point cloud set and the reverse scan point cloud set.

[0009] The embodiments of this application can segment the point cloud into a forward scan point cloud set and a reverse scan point cloud set corresponding to the target plane from the point cloud, thereby enabling the thickness of the target plane to be determined quickly and accurately using the forward scan point cloud set and the reverse scan point cloud set.

[0010] In some embodiments, controlling the lidar to perform point cloud correction based on the target gain coefficient includes: acquiring a preset correction model, the correction model being used to compensate for the pitch deviation of the lidar caused by the change in the direction of the slow-axis drive current, the slow-axis drive current being used to drive the galvanometer device to rotate around the slow axis; inputting the target gain coefficient into the correction model to obtain a target pitch compensation value; acquiring a point cloud compensation model, the point cloud compensation model being used to correct the point cloud deviation of the target point cloud caused by the pitch deviation; inputting the target pitch compensation value into the point cloud compensation model to correct the target point cloud, thereby obtaining a corrected target point cloud.

[0011] This application embodiment, by setting a correction model and a point cloud compensation model, can reliably and accurately correct the target point cloud using the calibrated target gain coefficient, thereby improving the accuracy and stability of the target point cloud and facilitating reliable and accurate object detection by lidar.

[0012] In some embodiments, obtaining a preset calibration model includes: obtaining a fast axis feedback signal; generating an ideal slow axis feedback signal based on the fast axis feedback signal; obtaining a real slow axis feedback signal, wherein the real slow axis feedback signal, the fast axis feedback signal, and the ideal slow axis feedback signal are all associated with the gain coefficient; and generating a calibration model based on the ideal slow axis feedback signal and the real slow axis feedback signal.

[0013] This application embodiment uses fast-axis feedback signal, ideal slow-axis feedback signal and real slow-axis feedback signal to construct a correction model. The correction model can accurately present the effect of the slow-axis feedback signal when the reverse of the slow-axis drive current changes. This is beneficial to obtaining accurate and reliable target pitch compensation value, and thus can help the lidar to accurately and reliably correct the target point cloud.

[0014] In some embodiments, generating a correction model based on the ideal slow-axis feedback signal and the actual slow-axis feedback signal includes: subtracting the ideal slow-axis feedback signal from the actual slow-axis feedback signal to obtain a slow-axis feedback difference; and generating a correction model based on the slow-axis feedback difference.

[0015] In this embodiment, the difference between the real slow-axis feedback signal and the ideal slow-axis feedback signal is used to obtain the slow-axis feedback difference, which can reflect other abnormal influences on the slow-axis feedback signal, including the influence of the slow-axis drive current changing in reverse. Therefore, the target pitch compensation value output by the correction model generated based on the slow-axis feedback difference can accurately correct the point cloud distortion on the slow-axis scan.

[0016] In some embodiments, obtaining the point cloud compensation model includes: obtaining a rotation matrix corresponding to the pitch direction, wherein the rotation matrix is ​​a matrix about the pitch compensation value; obtaining the laser emission vector of the lidar; and generating a point cloud compensation model based on the rotation matrix and the laser emission vector.

[0017] Point cloud distortion in the slow-axis scanning direction can be corrected by adjusting the point cloud position in the pitch direction. Therefore, this application embodiment generates a point cloud compensation model based on the rotation matrix and laser emission vector corresponding to the pitch direction. This point cloud compensation model can reliably and accurately correct point cloud distortion in the slow-axis scanning direction.

[0018] In some embodiments, obtaining the point cloud dataset obtained by the lidar scanning the target plane in both directions includes: controlling the lidar to scan the target plane in both directions according to a preset gain step value to obtain point clouds; and combining all point clouds to obtain a point cloud dataset.

[0019] This application embodiment obtains the point cloud for each frame by gradually increasing or decreasing the gain coefficient according to the gain step value. The gain step value is small, so the thickness of the point cloud between each two adjacent frames does not change much. This allows the thickness of all point clouds to be close or linearly change. In this way, the minimum thickness and the target gain coefficient corresponding to the minimum thickness can be found with high resolution and fine granularity. The target gain coefficient is relatively accurate and can compensate for the deviation that exists when the lidar performs forward and reverse scanning to the greatest extent, thereby improving the accuracy of the lidar.

[0020] In a second aspect, embodiments of this application provide a computer device including a memory and a processor. The memory is connected to the processor, and the processor is configured to execute one or more computer programs stored in the memory. When the processor executes the one or more computer programs, it causes the computer device to implement the point cloud correction method described above.

[0021] In a third aspect, embodiments of this application provide a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a processor, cause the processor to perform the point cloud correction method described above. Attached Figure Description

[0022] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a schematic diagram of the circuit structure of a galvanometer device provided in an embodiment of this application; Figure 2 A waveform diagram of a slow-axis drive current provided for an embodiment of this application; Figure 3 A schematic diagram showing the layering phenomenon on the ground when the lidar provided in the embodiment of this application performs forward and reverse scanning without correcting the point cloud; Figure 4 A schematic flowchart of a point cloud correction method provided in an embodiment of this application; Figure 5 A schematic diagram of a scenario for calibrating the target gain coefficient provided in an embodiment of this application; Figure 6 A schematic flowchart for calibrating the target gain coefficient provided in an embodiment of this application; Figure 7 The illustrations show the effects of the ground point cloud without correction and after correction, as provided in the embodiments of this application. Figure 8 This is a schematic diagram of the structure of a point cloud correction device provided in an embodiment of this application; Figure 9 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0025] It should be noted that, unless there is a conflict, the various features in the embodiments of this application can be combined with each other, all of which are within the protection scope of this application. Furthermore, although functional modules are divided in the device schematic diagram and a logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than the module division in the device or the order in the flowchart. Moreover, the terms "first," "second," and "third" used in this application do not limit the data or execution order, but only distinguish identical or similar items with essentially the same function and effect.

[0026] LiDAR (LiDAR) detects target position, velocity, and other characteristic information by emitting laser beams. Typically, a LiDAR system includes a galvanometer to control the scanning direction of the laser beam to scan the object. The galvanometer reflects the laser beam through a mirror that moves simultaneously in two directions, thus achieving laser scanning of the object. The central axes corresponding to the two directions of mirror movement are called the slow axis and the fast axis, respectively. The mirror moves slowly around the slow axis and quickly around the fast axis.

[0027] The following embodiments of this application provide a galvanometer device. Please refer to... Figure 1 The galvanometer device 100 includes a reflector 11, a slow axis coil 12, a slow axis drive circuit 13, a slow axis feedback circuit 14, a fast axis coil 15, a fast axis drive circuit 16, a fast axis feedback circuit 17, and a controller 18.

[0028] The reflector 11 is placed in the magnetic field of the permanent magnet to receive the laser and moves around the slow axis and fast axis to change the laser emission direction and achieve scanning coverage of the target area.

[0029] The slow-axis coil 12 is distributed around a vertical torsion axis. When energized, the torque generated drives the reflector 11 to rotate around the slow axis, causing the laser to move in the vertical direction. Specifically, the slow-axis coil 12 can drive the reflector 11 to perform low-frequency, large-angle deflection motion around the slow axis.

[0030] The slow-axis drive circuit 13 is electrically connected to both the slow-axis coil 12 and the controller 18. It receives slow-axis angle commands from the controller 18 and transmits slow-axis drive current to the slow-axis coil 12. The energized slow-axis coil 12 generates a low-frequency, stable torque in the magnetic field. This torque drives the reflector 11 to rotate around the slow axis, adjusting its slow-axis deflection angle and speed, thus enabling the laser to move vertically. The slow-axis drive current is a low-frequency drive current, and its shape can be either sawtooth or stepped.

[0031] The slow axis feedback circuit 14 is electrically connected to the controller 18. It is used to acquire the actual slow axis deflection angle of the reflector 11 and convert the actual slow axis deflection angle into an electrical signal to feed back to the controller 18 so that the controller 18 can form a closed-loop control about the slow axis.

[0032] The fast-axis coil 15 is distributed around a horizontal torsion axis, thus it is orthogonally distributed in space with the slow-axis coil 12. The torque generated when the fast-axis coil 15 is energized drives the reflector 11 to rotate around the fast axis, causing the laser to move in the horizontal direction. The fast-axis coil 15 can drive the reflector 11 to reciprocate around the fast axis at a high frequency and a small angle.

[0033] The fast-axis drive circuit 16 is electrically connected to both the fast-axis coil 15 and the controller 18. It receives fast-axis angle commands from the controller 18 and transmits fast-axis drive current to the fast-axis coil 15. The energized fast-axis coil 15 generates a high-frequency alternating torque in the magnetic field. This high-frequency alternating torque drives the reflector 11 to rotate around the fast axis, adjusting the fast-axis deflection angle and speed of the reflector 11, thereby enabling the laser to move horizontally. The fast-axis drive current is a high-frequency drive current, which includes a sinusoidal current.

[0034] The fast-axis feedback circuit 17 is electrically connected to the controller 18. It is used to acquire the actual fast-axis deflection angle of the reflector 11 and convert the actual fast-axis deflection angle into an electrical signal to feed back to the controller 18, so that the controller 18 can form a closed-loop control about the fast axis. The fast-axis feedback circuit 17 and the slow-axis feedback circuit 14 can be capacitive feedback circuits, photoelectric feedback circuits, etc.

[0035] Under the joint control of the slow-axis coil 12 and the fast-axis coil 15, the reflector 11 controls the laser to scan in the horizontal and vertical directions, thereby realizing planar scanning.

[0036] Please see Figure 2The slow-axis drive current can be a sawtooth wave current. When the direction of the slow-axis drive current changes, such as at position point A (peak) or position point B (valley), the rate of change of the slow-axis drive current is large, which can easily lead to the generation of abnormal high-frequency signals in the magnetic field. This abnormal high-frequency signal is collected by the slow-axis feedback circuit 14. At the same time, the slow-axis feedback circuit 14 is also affected by the high-frequency influence of the fast-axis drive current. The signals corresponding to the above factors are superimposed with the actual slow-axis deflection angle and fed back to the controller 18, causing the slow-axis deflection angle calculated by the controller 18 to produce an error with the actual slow-axis deflection angle. This error will cause the point cloud of the lidar to be distorted in the scanning direction of the slow axis, thereby affecting the performance of the lidar.

[0037] Related technologies use filtering circuits to filter out abnormal high-frequency signals in order to extract signals corresponding to the true slow axis deflection angle. However, the addition of filtering circuits will cause the slow axis feedback circuit 14 to introduce a larger phase delay, resulting in lag in the closed-loop control of the controller 18 about the slow axis, which in turn affects the quality of the point cloud.

[0038] Please see Figure 3 Due to the influence of the changing direction of the slow-axis drive current and the high frequency of the fast-axis drive current, a layering phenomenon easily occurs when the lidar scans the ground. Specifically, when the lidar scans the ground forward, the ground position resolved from the partial point cloud obtained from the forward scan is a straight line L1; when the lidar scans the ground backward, the ground position resolved from the partial point cloud obtained from the backward scan is a straight line L2, while the actual ground position is a straight line L0. Therefore, due to the influence of the changing direction of the slow-axis drive current, when the lidar scans the ground, one ground surface is resolved into two, resulting in ground layering.

[0039] Therefore, the point cloud correction method provided in this application can correct the point cloud and compensate for the error in the vertical direction, so that the lidar can obtain a consistent plane when scanning the same plane in both directions without producing layering. This is beneficial to improving the accuracy and reliability of lidar in detecting objects.

[0040] The following embodiments of this application provide a point cloud correction method applied to a lidar system, which includes a galvanometer device. Please refer to... Figure 4 The point cloud correction method is implemented through steps S41 to S43 in this embodiment of the application, as detailed below: Step S41: Obtain the point cloud dataset of the target plane scanned by the lidar in both directions.

[0041] The target plane can be any type of plane; for example, it could be the ground or a large desktop. Forward and reverse scanning include forward scanning and reverse scanning. Forward scanning involves the lidar-driven galvanometer device controlling the laser to scan along a first horizontal direction. Reverse scanning involves the lidar-driven galvanometer device controlling the laser to scan along a second horizontal direction, which is opposite to the first horizontal direction. For example, forward scanning is a horizontal scan to the right, and reverse scanning is a horizontal scan to the left.

[0042] The point cloud dataset is a collection of point clouds obtained by a LiDAR through multiple forward and reverse scans. The dataset consists of multiple frames of point clouds, each obtained by the LiDAR using a corresponding gain coefficient to drive a galvanometer device to scan the target plane around the fast and slow axes. The gain coefficient adjusts the fast-axis or slow-axis drive current, and is directly proportional to the absolute value of the gain coefficient. A larger absolute value of the gain coefficient results in a larger fast-axis or slow-axis drive current, and vice versa.

[0043] For example, the lidar uses a gain factor A1 to control the galvanometer to scan the target plane around the fast and slow axes respectively, obtaining point cloud D1. The lidar uses a gain factor A2 to control the galvanometer to scan the target plane around the fast and slow axes respectively, obtaining point cloud D2. The lidar uses a gain factor A3 to control the galvanometer to scan the target plane around the fast and slow axes respectively, obtaining point cloud D3. And so on, which will not be elaborated here.

[0044] In some embodiments, to improve the acquisition of point clouds corresponding to different fine-grained gain coefficients, this application embodiment controls the lidar to perform forward and reverse scanning of the target plane according to a preset gain step value, so as to obtain point clouds under different gain coefficients. Specifically, obtaining the point cloud dataset of the lidar scanning the target plane in both directions includes the following steps: controlling the lidar to perform forward and reverse scanning of the target plane according to a preset gain step value to obtain point clouds, and combining all point clouds to obtain a point cloud dataset.

[0045] The gain step value is customized by the designer based on engineering experience; for example, the gain step value is 0.01. This application embodiment obtains the gain coefficient range and the number of acquisitions. The gain coefficient range is jointly defined by the minimum gain coefficient and the maximum gain coefficient. Based on the minimum gain coefficient, the gain step value, and the number of acquisitions within the gain coefficient range, the current gain coefficient is determined. When the current gain coefficient is less than or equal to the maximum gain coefficient, the lidar is controlled to perform forward and reverse scanning of the target plane based on the current gain coefficient to obtain a point cloud. When the current gain coefficient is greater than the maximum gain coefficient, the forward and reverse scanning of the target plane by the lidar is stopped.

[0046] For example, the gain coefficient ranges from [-0.1, 0.1], the minimum gain coefficient is -0.1, the maximum gain coefficient is 0.1, the gain step value is 0.01, and the expression for the current gain coefficient is: When the number of samplings is 1, the current gain coefficient is... The lidar is configured with either a fast-axis drive current or a slow-axis drive current according to the current gain coefficient of -0.1, causing the galvanometer to move around the fast and slow axes to scan the target plane, thereby obtaining a point cloud corresponding to the current gain coefficient of -0.1. When the number of acquisitions is 2, the current gain coefficient is... The lidar is configured with either a fast-axis drive current or a slow-axis drive current according to the current gain coefficient of -0.1, causing the galvanometer to move around the fast and slow axes to scan the target plane, thereby obtaining a point cloud corresponding to the current gain coefficient of -0.09. When the number of acquisitions is 12, the current gain coefficient is... The lidar is configured with either a fast-axis drive current or a slow-axis drive current according to the current gain coefficient of 0.01, causing the galvanometer to move around the fast and slow axes to scan the target plane, thereby obtaining a point cloud corresponding to the current gain coefficient of 0.01. And so on.

[0047] Step S42: Determine the target gain coefficient based on the point cloud dataset.

[0048] The target gain coefficient is used to control the distortion of the LiDAR correction point cloud in the slow-axis scanning direction, thus minimizing the deviation between the corrected forward scan point cloud and the corrected reverse scan point cloud about the same target plane. The forward scan point cloud is a portion of the point cloud collected by the LiDAR during forward scanning, and the reverse scan point cloud is a portion of the point cloud collected by the LiDAR during reverse scanning. It can be understood that the forward scan point cloud and the reverse scan point cloud appear in the same frame of point cloud.

[0049] The slow-axis scanning direction is vertical. Minimum deviation refers to the minimum difference between the first position and the second position. The first position is the position of the target plane obtained by resolving a portion of the point cloud obtained from the forward scan, and the second position is the position of the target plane obtained by resolving a portion of the point cloud obtained from the reverse scan.

[0050] Understandably, the minimum deviation can be 0 or within a specific industry-acceptable range. When the minimum deviation is 0, it means that after the LiDAR corrects the point cloud according to the target gain coefficient, the first position and the second position coincide. That is, the target gain coefficient is very accurate. After the LiDAR corrects the forward and reverse scanning point clouds using the target gain coefficient, the first position of the target plane resolved from the corrected forward scanning point cloud and the second position of the target plane resolved from the corrected reverse scanning point cloud are consistent. When the minimum deviation is not 0 but within the specific industry-acceptable range, considering the inherent error, even after the LiDAR corrects the forward and reverse scanning point clouds using the target gain coefficient, the first and second positions are still not completely equal. However, the difference between the first and second positions is within the specific industry-acceptable range. It can be approximately considered that the corrected forward scanning point cloud and the corrected reverse scanning point cloud are approximately consistent. That is, this situation can be regarded as the target plane not exhibiting layering.

[0051] In this embodiment of the application, the target gain coefficient is determined based on the point cloud dataset through steps S421 to S423, as detailed below: Step S421: Determine the thickness of the target plane based on the point cloud.

[0052] The thickness is the height of the target plane in the vertical direction. Determining the thickness of the target plane based on the point cloud includes the following steps: extracting the forward scan point cloud set corresponding to the target plane from the point cloud; extracting the reverse scan point cloud set corresponding to the target plane from the point cloud; and determining the thickness of the target plane based on the forward scan point cloud set and the reverse scan point cloud set. In this embodiment, the forward scan point cloud set and the reverse scan point cloud set corresponding to the target plane are segmented from the point cloud, thereby enabling the rapid and accurate determination of the target plane's thickness using these two point cloud sets.

[0053] Each point cloud point is configured with coordinates (x, y, z) in the lidar coordinate system. Extracting the forward scan point cloud set and the reverse scan point cloud set corresponding to the target plane from the point cloud includes the following steps: extracting the forward scan point cloud and the reverse scan point cloud from the point cloud; performing spatial matching processing on the forward scan point cloud and the reverse scan point cloud based on the point cloud matching algorithm to obtain the matched point cloud; and extracting the forward scan point cloud set and the reverse scan point cloud set corresponding to the target plane from the matched point cloud.

[0054] The point cloud matching algorithm can be the ICP matching algorithm. After spatial matching processing, the x-axis and y-axis coordinates of the first point cloud in the forward-scanning point cloud can coincide with the x-axis and y-axis coordinates of the corresponding second point cloud in the reverse-scanning point cloud. However, the z-axis coordinates of the first point cloud and the second point cloud are different. In this paper, (x-axis coordinate, y-axis coordinate) are defined as horizontal coordinates.

[0055] The matched point cloud includes the matched forward scan point cloud and the matched reverse scan point cloud. Extracting the forward scan point cloud set and the reverse scan point cloud set corresponding to the target plane from the matched point cloud includes the following steps: Based on the plane segmentation algorithm, extract the forward scan point cloud set corresponding to the target plane from the matched forward scan point cloud, and extract the reverse scan point cloud set corresponding to the target plane from the matched reverse scan point cloud. Perform downsampling and noise reduction processing on the forward scan point cloud set and the reverse scan point cloud set.

[0056] The forward scan point cloud set includes multiple forward scan point cloud points, and the reverse scan point cloud set includes multiple reverse scan point cloud points. The forward scan point cloud points are point cloud points corresponding to the target plane obtained based on the forward scanning method, and the reverse scan point cloud points are point cloud points corresponding to the target plane obtained based on the reverse scanning method.

[0057] Determining the thickness of a target plane based on the forward scan point cloud set and the reverse scan point cloud set includes the following steps: extracting a target forward scan point cloud from the forward scan point cloud set, and extracting the target reverse scan point cloud corresponding to the target forward scan point cloud from the reverse scan point cloud set, wherein the horizontal coordinates of the target forward scan point cloud and the target reverse scan point cloud are the same; calculating the absolute value of the difference between the z-axis coordinates of the target forward scan point cloud and the target reverse scan point cloud; and determining the thickness of the target plane based on the absolute value of the differences corresponding to all target forward scan point cloud sets.

[0058] In this embodiment, the absolute value of the difference between the z-axis coordinate of the target forward scan point cloud and the z-axis coordinate of the target reverse scan point cloud is calculated according to the following formula, as shown below: Formula 1 Wherein, the coordinates of the target forward scan point cloud are The coordinates of the target backscan point cloud are , Let z be the z-axis coordinate of the target positive scan point cloud point. Let be the z-axis coordinate of the target backscanned point cloud point, and be the absolute value of the difference. Wherein, and equal.

[0059] In some embodiments, determining the thickness of the target plane based on the absolute value of the difference between all target forward scan point cloud points includes the following steps: adding the absolute values ​​of the difference between all target forward scan point cloud points to obtain the absolute value of the total difference; dividing the absolute value of the total difference by the total number of points in all target forward scan point cloud points to obtain the average difference; and using the average difference as the thickness of the target plane.

[0060] In other embodiments, determining the thickness of the target plane based on the absolute value of the differences corresponding to all target forward scan point cloud points includes the following steps: finding the minimum absolute value among the absolute values ​​of the differences corresponding to all target forward scan point cloud points, and using the minimum absolute value as the thickness of the target plane.

[0061] Step S422: Determine the minimum thickness among the thicknesses of the multi-frame point cloud.

[0062] In this embodiment, the thicknesses of multiple point clouds are sorted according to a preset sorting algorithm to obtain a sorting result, and the minimum thickness is determined based on the sorting result. When the preset sorting algorithm is a descending sorting algorithm, the thickness at the end of the sorting result is selected as the minimum thickness. When the preset sorting algorithm is a ascending sorting algorithm, the thickness at the beginning of the sorting result is selected as the minimum thickness.

[0063] Step S423: Determine the gain coefficient corresponding to the minimum thickness as the target gain coefficient.

[0064] Each frame of point cloud is obtained by a lidar using a corresponding gain coefficient to drive a galvanometer device to scan the target plane around the fast and slow axes respectively. Therefore, the attribute expression of the point cloud is (Di, Ai), where Di is the point cloud of the i-th frame and Ai is the gain coefficient corresponding to the point cloud of the i-th frame. After the embodiment of this application determines the thickness corresponding to each frame of point cloud based on the above method, the attribute expression of the point cloud can be updated to (Di, Ai, Hi), where Hi is the thickness corresponding to the point cloud of the i-th frame. The total attribute expression of all point clouds for the same target plane is {(D1, A1, H1), (D2, A2, H2), (D3, A3, H3), ..., (Di, Ai, Hi), ..., (Dn, An, Hn)}.

[0065] When the minimum thickness Hmin is obtained in this embodiment, the embodiment uses the minimum thickness Hmin as an index to search for the attribute expression of the point cloud with a thickness equal to the minimum thickness Hmin from the total attribute expression of all point clouds about the same target plane, and selects the gain coefficient contained in the attribute expression of that point cloud as the target gain coefficient. For example, if the thickness of the 9th point cloud is equal to the minimum thickness Hmin, then the gain coefficient contained in the attribute expression of the 9th point cloud is set as the target gain coefficient.

[0066] It is understandable that the minimum thickness reflects the minimum deviation between forward and reverse scanning. In this application embodiment, the gain coefficient corresponding to the minimum thickness is used as the target gain coefficient. The lidar is controlled to correct the point cloud with low deviation and high accuracy according to the target gain coefficient, which is beneficial to improving the solution accuracy of the point cloud.

[0067] It is also understood that the embodiments of this application obtain each frame of point cloud by gradually increasing or decreasing the gain coefficient according to the gain step value. The gain step value is small, so that the thickness of each two adjacent frames of point cloud does not change much. This makes the thickness of all point clouds close or linear. In this way, the minimum thickness and the target gain coefficient corresponding to the minimum thickness can be found with high resolution and high granularity. The target gain coefficient is relatively accurate and can compensate for the deviation of the lidar when performing forward and reverse scanning to the greatest extent, thereby improving the accuracy of the lidar.

[0068] To obtain the target gain coefficient corresponding to the minimum thickness, embodiments of this application provide a calibration process for calibrating the target gain coefficient. Embodiments of this application combine... Figure 5 and Figure 6 The calibration process is explained below: S61. Construct a calibration site 500, which is required to be 7m wide and 17m long. A lidar 51 is installed at the calibration site, located at the center of the left edge of the calibration site 500.

[0069] S62, set the gain coefficient range to [-0.1, 0.1].

[0070] S63 determines the current gain coefficient for each forward and reverse scan operation based on a gain step value of 0.01 and the number of acquisitions.

[0071] S64, based on the current gain coefficient, configure the fast axis drive current or the slow axis drive current, so that the galvanometer device moves around the fast axis and the slow axis to scan the ground 52, thereby obtaining each frame of point cloud.

[0072] S65 extracts the forward scan point cloud set and the reverse scan point cloud set about the ground from the current point cloud.

[0073] S66, determine the ground thickness corresponding to the current point cloud based on the forward scan point cloud set and the reverse scan point cloud set.

[0074] S67, find the minimum thickness from the ground thickness of all point clouds.

[0075] S68, determine the gain coefficient with minimum thickness as the target gain coefficient.

[0076] S69, writes the target gain coefficient into the lidar.

[0077] Based on the above calibration process, the target gain coefficient is obtained in this embodiment. By calibrating the target gain coefficient, this embodiment helps to correct the distortion that easily occurs in the slow-axis scanning direction of the point cloud, improves the solution accuracy of the point cloud, solves the problem of point cloud quality reduction caused by device feedback errors, and improves the accuracy and stability of the point cloud.

[0078] Step S43: Control the lidar to perform point cloud correction operation based on the target gain coefficient.

[0079] The point cloud correction operation corrects the target point cloud based on the target gain coefficient, making the corrected target point cloud close to or consistent with the real point cloud. In this embodiment, steps S431 to S434 control the lidar to perform the point cloud correction operation based on the target gain coefficient, as detailed below: Step S431: Obtain the preset calibration model.

[0080] The calibration model is used to compensate for the pitch deviation of the lidar caused by the change in the direction of the slow axis drive current, which is used to drive the galvanometer device to rotate around the slow axis.

[0081] Obtaining the preset calibration model includes the following steps: obtaining the fast axis feedback signal, generating an ideal slow axis feedback signal based on the fast axis feedback signal, obtaining the real slow axis feedback signal, wherein the real slow axis feedback signal, the fast axis feedback signal, and the ideal slow axis feedback signal are all related to the gain coefficient, and generating a calibration model based on the ideal slow axis feedback signal and the real slow axis feedback signal.

[0082] The fast-axis feedback signal is the signal acquired by the fast-axis feedback circuit of the lidar. The expression for the fast-axis feedback signal is shown in Equation 2: Formula 2 in, Here, A is the fast-axis feedback signal, f is the gain coefficient, t is the frequency, and phase1 is the first lag phase. The frequency f can be obtained directly.

[0083] From equation two, we can obtain: Formula 3 The ideal slow-axis feedback signal is the signal acquired by the slow-axis feedback circuit of the lidar when it is not affected by changes in the direction of the slow-axis drive current. The expression for the ideal slow-axis feedback signal is shown in Equation 4: Formula 4 in, The ideal slow-axis feedback signal is given, and K is a user-defined feedback coefficient. This is the second lagging phase.

[0084] The true slow-axis feedback signal is the signal collected by the slow-axis feedback circuit of the lidar under actual conditions. It can be understood that the true slow-axis feedback signal is composed of a linear signal and a sinusoidal signal superimposed on each other. The current error affects the sinusoidal signal in the true slow-axis feedback signal. The sinusoidal signal is coupled to the slow-axis feedback circuit by the influence of the fast-axis drive current of the sinusoidal wave. Therefore, the frequency of the true slow-axis feedback signal is the same as the frequency of the fast-axis feedback signal.

[0085] The expression for the actual slow-axis feedback signal is shown in Equation 5: Formula 5 in, This is a true slow-axis feedback signal. It is the third lagging phase.

[0086] The generation of a correction model based on the ideal slow-axis feedback signal and the actual slow-axis feedback signal includes the following steps: subtracting the ideal slow-axis feedback signal from the actual slow-axis feedback signal to obtain the slow-axis feedback difference, and generating a correction model based on the slow-axis feedback difference. In this embodiment, the slow-axis feedback difference obtained by subtracting the ideal slow-axis feedback signal from the actual slow-axis feedback signal can reflect other abnormal influences on the slow-axis feedback signal, including the effect of changes in the reverse direction of the slow-axis drive current. Therefore, the target pitch compensation value output by the correction model generated based on the slow-axis feedback difference can accurately correct point cloud distortion on the slow-axis scan.

[0087] The embodiments of this application use the following formula to obtain the slow axis feedback difference, as shown below: Formula Six set up: Formula 7 Formula 8 Based on equations six, seven, and eight, we have: Formula Nine Simplify equation nine. ; A=15, K=1 / 30; have: Formula 10 Formula 11 Based on equations six, ten, and eleven, we have: Formula Twelve Formula Thirteen Since Phase 3 is actually quite small, therefore, this item It can be discarded.

[0088] Therefore, the final correction model is: Formula Fourteen This application embodiment uses fast-axis feedback signal, ideal slow-axis feedback signal and real slow-axis feedback signal to construct a correction model. The correction model can accurately present the effect of the slow-axis feedback signal when the reverse of the slow-axis drive current changes. This is beneficial to obtaining accurate and reliable target pitch compensation value, and thus can help the lidar to accurately and reliably correct the target point cloud.

[0089] Step S432: Input the target gain coefficient into the correction model to obtain the target pitch compensation value.

[0090] In this embodiment, the target gain coefficient A' is input into the calibration model to obtain the target pitch compensation value Pitch_Calibration'. The target pitch compensation value is used to compensate for the point cloud deviation caused by pitch deviation. The target point cloud is the point cloud collected by the lidar after obtaining the target gain coefficient.

[0091] Step S433: Obtain the point cloud compensation model.

[0092] A point cloud compensation model is used to correct point cloud deviations caused by pitch deviations. Obtaining the point cloud compensation model includes the following steps: obtaining a rotation matrix corresponding to the pitch direction, where the rotation matrix is ​​a matrix about the pitch compensation value; obtaining the laser emission vector of the lidar; and generating the point cloud compensation model based on the rotation matrix and the laser emission vector. Point cloud distortion in the slow-axis scanning direction can be corrected by adjusting the point cloud position in the pitch direction. Therefore, this embodiment of the application generates a point cloud compensation model based on the rotation matrix corresponding to the pitch direction and the laser emission vector. This point cloud compensation model can reliably and accurately correct point cloud distortion in the slow-axis scanning direction.

[0093] The expression for the point cloud compensation model is: Formula 15 in, The target point cloud after correction. The rotation matrix chosen about the Y-axis. This is the laser emission vector.

[0094] This application embodiment, by setting a correction model and a point cloud compensation model, can reliably and accurately correct the target point cloud using the calibrated target gain coefficient, thereby improving the accuracy and stability of the target point cloud and facilitating reliable and accurate object detection by lidar.

[0095] Step S434: Input the target pitch compensation value into the point cloud compensation model to correct the target point cloud and obtain the corrected target point cloud.

[0096] In this embodiment, the target pitch compensation value 'Pitch_Calibration' is substituted into the point cloud compensation model shown in Equation 15 to obtain the corrected target point cloud. .

[0097] Please see Figure 7 The red point cloud represents the uncorrected ground point cloud, which exhibits layering, consisting of an upper ground point cloud 71 and a lower ground point cloud 72. After calibrating the target gain coefficient and correcting the upper ground point cloud 71 and lower ground point cloud 72 based on the target gain coefficient, the green ground point cloud 73 is obtained. Figure 7 As can be seen, the green ground point cloud 73 is relatively concentrated and does not show obvious layering. Therefore, the embodiment of this application can determine the target gain coefficient based on the point cloud dataset obtained by forward and reverse scanning, so that the lidar can use the target gain coefficient to correct the point cloud distortion that is prone to occur in the slow axis scanning direction, improve the point cloud layering situation during forward and reverse laser scanning, thus improving the point cloud solution accuracy, solving the problem of point cloud quality reduction caused by device feedback error, and improving the accuracy and stability of the point cloud.

[0098] It should be noted that in the above embodiments, there is no necessarily a certain order between the steps. Those skilled in the art can understand from the description of the embodiments of this application that the above steps may have different execution orders in different embodiments, that is, they may be executed in parallel or in turn, etc.

[0099] As another aspect of the embodiments of this application, this application provides a point cloud correction device. The point cloud correction device can be a software module, which includes several instructions stored in a memory. A processor can access the memory, call the instructions, and execute them to complete the point cloud correction methods described in the various embodiments above.

[0100] In some embodiments, the point cloud correction device can also be constructed from hardware devices. For example, the point cloud correction device can be constructed from one or more chips, which can work together to complete the point cloud correction methods described in the various embodiments above. As another example, the point cloud correction device can also be constructed from various logic devices, such as general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), microcontrollers, ARM (Acorn RISC Machine) or other programmable logic devices, discrete gate or transistor logic, discrete hardware components, or any combination of these components.

[0101] Please see Figure 8 The point cloud correction device 800 includes a point cloud acquisition module 81, a gain determination module 82, and a point cloud correction module 83.

[0102] The point cloud acquisition module 81 acquires a point cloud dataset from the forward and reverse scanning of the target plane by the lidar. The dataset includes multiple frames of point clouds, each obtained by the lidar using different gain coefficients to drive a galvanometer device to scan the target plane around the fast and slow axes respectively. The gain determination module 82 determines the target gain coefficient based on the point cloud dataset. The target gain coefficient controls the lidar to correct distortions in the point cloud along the slow-axis scanning direction. The point cloud correction module 83 controls the lidar to perform point cloud correction operations based on the target gain coefficient.

[0103] The embodiments of this application can determine the target gain coefficient based on the point cloud dataset obtained by forward and reverse scanning, so that the lidar can use the target gain coefficient to correct the distortion of the point cloud that is prone to occur in the slow axis scanning direction, improve the point cloud layering during forward and reverse laser scanning, thus improving the solution accuracy of the point cloud, solving the problem of point cloud quality reduction caused by device feedback errors, and improving the accuracy and stability of the point cloud.

[0104] In some embodiments, the gain determination module 82 is specifically used to: determine the thickness of the target plane based on the point cloud, determine the minimum thickness among the thicknesses of the point cloud in multiple frames, and determine the gain coefficient corresponding to the minimum thickness as the target gain coefficient.

[0105] In some embodiments, the gain determination module 82 is specifically used to: extract a set of forward-scanned point clouds corresponding to the target plane from the point cloud, extract a set of reverse-scanned point clouds corresponding to the target plane from the point cloud, and determine the thickness of the target plane based on the set of forward-scanned point clouds and the set of reverse-scanned point clouds.

[0106] In some embodiments, the point cloud correction module 83 is specifically used to: obtain a preset correction model, the correction model being used to compensate for the pitch deviation of the lidar caused by the change in the direction of the slow axis drive current, the slow axis drive current being used to drive the galvanometer device to rotate around the slow axis, input the target gain coefficient into the correction model to obtain the target pitch compensation value, obtain a point cloud compensation model, the point cloud compensation model being used to correct the point cloud deviation of the target point cloud caused by the pitch deviation, input the target pitch compensation value into the point cloud compensation model to correct the target point cloud, and obtain the corrected target point cloud.

[0107] In some embodiments, the point cloud correction module 83 is specifically used to: acquire a fast axis feedback signal, generate an ideal slow axis feedback signal based on the fast axis feedback signal, acquire a real slow axis feedback signal, wherein the real slow axis feedback signal, the fast axis feedback signal, and the ideal slow axis feedback signal are all related to the gain coefficient, and generate a correction model based on the ideal slow axis feedback signal and the real slow axis feedback signal.

[0108] In some embodiments, the point cloud correction module 83 is specifically used to: subtract the ideal slow axis feedback signal from the real slow axis feedback signal to obtain the slow axis feedback difference, and generate a correction model based on the slow axis feedback difference.

[0109] In some embodiments, the point cloud correction module 83 is specifically used to: obtain a rotation matrix corresponding to the pitch direction, wherein the rotation matrix is ​​a matrix about the pitch compensation value; obtain the laser emission vector of the lidar; and generate a point cloud compensation model based on the rotation matrix and the laser emission vector.

[0110] In some embodiments, the point cloud correction module 83 is specifically used to: control the lidar to perform forward and reverse scanning of the target plane according to a preset gain step value, obtain point clouds, and combine all point clouds to obtain a point cloud dataset.

[0111] It should be noted that the above-described point cloud correction device can execute the point cloud correction method provided in the embodiments of this application, and has the corresponding functional modules and beneficial effects of the method. Technical details not described in detail in the embodiments of the point cloud correction device can be found in the point cloud correction method provided in the embodiments of this application.

[0112] See Figure 9 , Figure 9 This is a schematic diagram of a computer device provided in an embodiment of this application. The computer device 900 includes one or more processors 91 and a memory 92. The memory 92 is connected to one or more processors 91, for example, via a bus.

[0113] Processor 91 is configured to support the computer device in performing the corresponding functions in the methods described in the above method embodiments. The processor may be a central processing unit (CPU), a network processor (NP), a hardware chip, or any combination thereof. The aforementioned hardware chip may be an application-specific integrated circuit (ASIC), a programmable logic device (PLD), or a combination thereof. The aforementioned PLD may be a complex programmable logic device (CPLD), a field-programmable gate array (FPGA), a generic array logic (GAL), or any combination thereof.

[0114] Memory 92 is used to store program code, etc. Memory may include volatile memory (VM), such as random access memory (RAM); memory may also include non-volatile memory (NVM), such as read-only memory (ROM), flash memory, hard disk drive (HDD), or solid-state drive (SSD); memory may also include combinations of the above types of memory.

[0115] The memory 92 can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules, such as the program instructions / modules corresponding to the point cloud correction method in the embodiments of this application. The processor executes various functional applications and data processing of the point cloud correction method and the point cloud correction device by running the non-volatile software programs, instructions, and modules stored in the memory, that is, it realizes the functions of each module or unit of the point cloud correction method and the point cloud correction device provided in the above method embodiments.

[0116] The memory 92 may include a program storage area and a data storage area, wherein the program storage area may store the operating system and applications required for at least one function. The data storage area may store data created based on the use of the point cloud correction device, etc. In some embodiments, the memory may optionally include memory remotely configured relative to the processor, which can be connected to the point cloud correction device via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0117] The one or more modules are stored in the memory. When executed by the one or more processors, they perform the point cloud correction method in any of the above method embodiments. For example, they perform the method steps described in the above method embodiments to realize the functions of the modules described in the above device embodiments.

[0118] This application also provides a computer-readable storage medium storing a computer program, the computer program including program instructions, which, when executed by a computer device, cause the computer device to perform the method described in the foregoing embodiments.

[0119] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0120] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. A point cloud correction method applied to a lidar system, the lidar system comprising a galvanometer device, characterized in that, The point cloud correction method includes: The point cloud dataset obtained by the lidar scanning the target plane in both directions includes multiple frames of point cloud. Each frame of point cloud is obtained by the lidar driving the galvanometer device to scan the target plane around the fast axis and the slow axis respectively using different gain coefficients. The target gain coefficient is determined based on the point cloud dataset, and the target gain coefficient is used to control the distortion that occurs in the slow axis scanning direction of the lidar correction point cloud. The lidar is controlled to perform point cloud correction operations based on the target gain coefficient.

2. The point cloud correction method according to claim 1, characterized in that, Determining the target gain coefficient based on the point cloud dataset includes: The thickness of the target plane is determined based on the point cloud; Determine the minimum thickness among the thicknesses of the point cloud across multiple frames; The gain coefficient corresponding to the minimum thickness is determined as the target gain coefficient.

3. The point cloud correction method according to claim 2, characterized in that, Determining the thickness of the target plane based on the point cloud includes: Extract the set of forward-scanned point clouds corresponding to the target plane from the point cloud; Extract the set of anti-scanned point clouds corresponding to the target plane from the point cloud; The thickness of the target plane is determined based on the forward scan point cloud set and the reverse scan point cloud set.

4. The point cloud correction method according to claim 1, characterized in that, The step of controlling the lidar to perform point cloud correction based on the target gain coefficient includes: A preset calibration model is obtained, which is used to compensate for the pitch deviation of the lidar caused by the change in the direction of the slow axis drive current. The slow axis drive current is used to drive the galvanometer device to rotate around the slow axis. The target gain coefficient is input into the correction model to obtain the target pitch compensation value; A point cloud compensation model is obtained, which is used to correct the point cloud deviation of the target point cloud caused by the pitch deviation. The target pitch compensation value is input into the point cloud compensation model to correct the target point cloud, thus obtaining the corrected target point cloud.

5. The point cloud correction method according to claim 4, characterized in that, The process of obtaining the preset correction model includes: Obtain the fast axis feedback signal; An ideal slow axis feedback signal is generated based on the fast axis feedback signal; Obtain the true slow axis feedback signal, wherein the true slow axis feedback signal, the fast axis feedback signal, and the ideal slow axis feedback signal are all related to the gain coefficient; A correction model is generated based on the ideal slow-axis feedback signal and the real slow-axis feedback signal.

6. The point cloud correction method according to claim 5, characterized in that, The generation of the correction model based on the ideal slow-axis feedback signal and the real slow-axis feedback signal includes: Subtract the ideal slow axis feedback signal from the actual slow axis feedback signal to obtain the slow axis feedback difference; A correction model is generated based on the slow axis feedback difference.

7. The point cloud correction method according to claim 4, characterized in that, The point cloud compensation model includes: Obtain the rotation matrix corresponding to the pitch direction, wherein the rotation matrix is ​​a matrix about the pitch compensation value; Obtain the laser emission vector of the lidar; A point cloud compensation model is generated based on the rotation matrix and the laser emission vector.

8. The point cloud correction method according to any one of claims 1 to 7, characterized in that, The step of acquiring the point cloud dataset obtained by the lidar scanning the target plane from both directions includes: According to the preset gain step value, the lidar is controlled to scan the target plane in both directions to obtain point cloud; Combine all point clouds to obtain a point cloud dataset.

9. A computer device, characterized in that, The device includes a memory and a processor, the memory being connected to the processor, the processor being configured to execute one or more computer programs stored in the memory, the processor causing the computer device to implement the point cloud correction method as described in any one of claims 1-8 when executing the one or more computer programs.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, the computer program including program instructions that, when executed by a processor, cause the processor to perform the point cloud correction method as described in any one of claims 1-8.

Images