System and method based on joint automatic calibration of multiple single-line laser radars

Abstract

The present invention discloses a system for joint automatic calibration based on multiple single-line laser radars. The system comprises a laser rangefinder for measuring the actual coordinates of the center point of a calibration object; laser radars located on the left and right sides of the calibration object respectively collect depth information of the calibration object; a point cloud acquisition module converts the collected depth information of the calibration object into point cloud data information; a point cloud rotation module obtains point cloud data information on the left side of each rotation angle and point cloud data information on the right side of each rotation angle through a rotation matrix; a point cloud processing module obtains multiple spliced ​​point clouds, generates clustered point clouds, selects an area of ​​interest of the calibration object for cropping, generates a cropped point cloud, and obtains predicted coordinates of the center point of the calibration object; calculates the error value between the predicted coordinates of the center point of the calibration object corresponding to each spliced ​​point cloud and the actual coordinates of the center point of the calibration object, and uses the corresponding left point cloud data rotation angle and right point cloud data rotation angle in the spliced ​​point cloud when the error value is minimized as the calibration correction angle.

Description

Technical Field

[0001] The present invention relates to the field of measurement technology, and in particular to a system and method based on the joint automatic calibration of multiple single-line laser radars. Background Art

[0002] When performing lidar position calibration, it is usually necessary to place a regular calibration object with known position information within the lidar measurement range, usually a rectangular calibration object, and then use the lidar to scan the calibration object and calibrate the lidar based on the position information of the calibration object.

[0003] Typically, a LiDAR scan is used to capture a dataset of points on the surface of a calibration object, forming a point cloud. Based on the positional information in the point cloud, the LiDAR's position relative to any origin can be determined. However, in the actual point cloud data collection process, LiDAR point clouds inevitably contain noise and outliers (point clouds with incorrect positional information, i.e., outliers) due to factors such as inherent errors in the acquisition equipment, deviations from the installation angle and set values, and the presence of debris such as small rocks in the measurement area.

[0004] These issues not only affect the accuracy of the point cloud model but also have a significant impact on subsequent LiDAR calibration. To address this issue, most researchers treat outliers as noise, remove them through algorithms such as filtering and denoising, and then manually adjust the LiDAR rotation angle to improve the accuracy of LiDAR data acquisition. However, this method is time-consuming and labor-intensive, and the processing accuracy is low. Therefore, subsequent processing of the point cloud collected by the LiDAR and correction of the LiDAR installation rotation angle will improve the speed and accuracy of LiDAR calibration. Summary of the Invention

[0005] The purpose of the present invention is to provide a system and method based on the joint automatic calibration of multiple single-line laser radars, which improves the speed and accuracy of laser radar calibration.

[0006] To achieve this purpose, the present invention designs a system based on the joint automatic calibration of multiple single-line laser radars, which includes:

[0007] The laser rangefinder is used to measure the actual coordinates of the center point of the calibration object;

[0008] The first laser radar located on the left side of the calibration object is used to collect the depth information of the calibration object on the left side of the calibration object, and the second laser radar located on the right side of the calibration object is used to collect the depth information of the calibration object on the right side of the calibration object;

[0009] The point cloud acquisition module is used to convert the depth information of the left calibration object into the left point cloud data information of the calibration object, and convert the depth information of the right calibration object into the right point cloud data information of the calibration object;

[0010] The point cloud rotation module is used to rotate the left point cloud data information N times within the preset rotation angle range through the rotation matrix to obtain N rotation angle left point cloud data information, and to rotate the right point cloud data information M times within the preset rotation angle range through the rotation matrix to obtain M rotation angle right point cloud data information;

[0011] The point cloud processing module is used to splice the point cloud data information on the left side of each rotation angle with the point cloud data information on the right side of each rotation angle to obtain corresponding multiple spliced ​​point clouds; cluster each spliced ​​point cloud to generate a clustered point cloud; crop the clustered point cloud according to the selected calibration object interest area to generate a cropped point cloud, and obtain the predicted coordinates of the calibration object center point based on the cropped point cloud; calculate the error value between the predicted coordinates of the calibration object center point corresponding to each spliced ​​point cloud and the actual coordinates of the calibration object center point, and use the corresponding left point cloud data rotation angle and right point cloud data rotation angle in the spliced ​​point cloud corresponding to the minimum error value as the calibration correction angle.

[0012] Beneficial effects of the present invention:

[0013] Compared with the traditional manual adjustment calibration method, the method of automatic calibration of multiple single-line lidars can significantly reduce the need and tediousness of manual operation, reduce human or installation errors, improve the consistency and repeatability of the calibration process, and reduce the time and manpower required for the calibration process, thereby reducing the overall cost. The automated calibration method can quickly complete the calibration of multiple lidars, allowing the system to be put into use more quickly or calibrated in a new application environment. Finally, the graphical interface is designed to improve the friendliness of non-professional users, reduce the technical details that users need to know during the calibration process, and make it easy for non-professionals to use. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a structural schematic diagram of the present invention;

[0015] Figure 2 is a flow chart of the present invention;

[0016] Figure 3 This is a schematic diagram of the point cloud information collected by the first laser radar;

[0017] Figure 4 This is a schematic diagram of the point cloud information collected by the second laser radar;

[0018] Figure 5 This is a schematic diagram after point cloud stitching;

[0019] Figure 6 Schematic diagram after clustering;

[0020] Figure 7 This is a schematic diagram after cropping;

[0021] Figure 8 Schematic diagram of the designed software interface;

[0022] Figure 9 A schematic diagram showing the result interface;

[0023] Figure 10 Schematic diagram of the point cloud collected by the adjusted lidar. DETAILED DESCRIPTION

[0024] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:

[0025] Example 1

[0026] A system based on the joint automatic calibration of multiple single-line laser radars, such as Figure 1 As shown, it includes:

[0027] The laser rangefinder is used to measure the actual coordinates of the center point of the calibration object. The specific design is to first place the calibration object in the interval scanned by each set of laser radars, and then use the laser rangefinder (or other high-precision equipment) to measure the actual coordinates (x0, y0) of the center point of each calibration object, so that it can be calibrated with the predicted coordinates of the calibration object center point predicted by the laser radar.

[0028] The first laser radar located on the left side of the calibration object is used to collect the depth information of the calibration object on the left side of the calibration object, and the second laser radar located on the right side of the calibration object is used to collect the depth information of the calibration object on the right side of the calibration object;

[0029] The point cloud acquisition module is used to convert the depth information of the left calibration object into the point cloud data information of the left calibration object (the laser radar starts from -45° and rotates 0.18° each time to collect information. Through trigonometric function conversion, the measured distance information is multiplied by the cosine and sine values ​​of the angle of rotation of the brushless motor in the laser radar at this time to obtain the coordinate value; through the loop, all the points scanned by the laser radar can be converted into coordinates) l ), convert the depth information of the right calibration object into the right point cloud data information of the calibration object (the right point cloud data information is P r ) (The point cloud data information is a point cloud data set, which is a data set of all points scanned by the laser radars on the left and right sides of the calibration object, that is, (x, y). This design is used to obtain the coordinates of the calibration object relative to the laser radars on the left and right sides;

[0030] The point cloud rotation module is used to rotate the left point cloud data information within the preset rotation angle range (the preset rotation angle range is (-5°, 5°), including -5° and 5°) N times (0.1° each time, 0≤N≤101) through the rotation matrix to obtain N rotation angle left point cloud data information, and to rotate the right point cloud data information within the preset rotation angle range M times (0.1° each time, 0≤M≤101) through the rotation matrix to obtain M rotation angle right point cloud data information (finally 101 2 point cloud rotation combination) (point cloud rotation, specifically, using a nested loop structure, the point cloud data information on the left side of the calibration object is rotated with the first laser radar position on the left side of the calibration object as the rotation axis, and then rotated in sequence within a certain range by a certain angle (from (-5°, 5°) each time 0.1°). Each rotation makes the point cloud data information on the right side of the calibration object take the second laser radar position on the right side of the calibration object as the rotation axis, and then rotated in sequence within a certain range by a certain angle (from (-5°, 5°) each time 0.1°). This design uses these two layers of nested loops to obtain the influence of the horizontal installation angle of the laser radars on the left and right sides of the calibration object on the point cloud acquisition, which facilitates the subsequent use of the results obtained from each rotation to predict the center point of the calibration object, thereby obtaining the calibration correction angles of the laser radars on the left and right sides of the calibration object;

[0031] The point cloud processing module is used to splice the point cloud data information on the left side of each rotation angle with the point cloud data information on the right side of each rotation angle to obtain corresponding multiple spliced ​​point clouds (P m ); clustering each spliced ​​point cloud to generate clustered point cloud (P d ); The clustered point cloud is clipped according to the selected calibration object interest area to generate the clipped point cloud (P c ), obtain the predicted coordinates of the center point of the calibration object according to the cropped point cloud; calculate the error value between the predicted coordinates of the center point of the calibration object corresponding to each spliced ​​point cloud and the actual coordinates of the center point of the calibration object, and use the corresponding left point cloud data rotation angle and right point cloud data rotation angle in the spliced ​​point cloud corresponding to the minimum error value as the calibration correction angle (during the operation of the point cloud rotation module, the point cloud data information collected by the first laser radar of the calibration object is rotated once every 0.1° in the range of -5° to 5° with the position of the first laser radar as the rotation axis. Each time it rotates, the point cloud data information collected by the second laser radar of the calibration object is rotated once every 0.1° in the range of -5° to 5° with the position of the second laser radar as the rotation axis. Then, the point cloud data information collected by the first laser radar and the second laser radar are multiplied by the rotation and translation matrix rotation_matrix_l = [cos(i), -sin(i), t x ; sin(i), cos(i), t y;0,0,1], rotation_matrix_r=[cos(j),-sin(j),t x ; sin(j), cos(j), t y ;0,0,1];After each rotation of the point cloud data information on the left and right sides of the calibration object, the point cloud data information on the left and the point cloud data information on the right are spliced ​​to obtain a spliced ​​point cloud, the spliced ​​point cloud is clustered to generate a clustered point cloud, the clustered point cloud is cropped according to the selected calibration object interest area to generate a cropped point cloud, the predicted coordinates of the center point of the calibration object are obtained according to the cropped point cloud, the error value between the predicted coordinates of the center point of the calibration object and the actual coordinates of the center point of the calibration object is calculated, the error value obtained by each rotation of the left and right laser radars to different angles is compared with the error value of the next angle, the smallest error value is retained, and then the smallest error value can be found until the two laser radars are rotated). This design is used to obtain the calibration correction angle of the first laser radar located on the left side of the calibration object and the second laser radar located on the right side of the calibration object, that is, the optimal angle in the horizontal position to compensate for the error caused by manually placing the laser radar, which can greatly improve the calibration accuracy of the laser radar.

[0032] In the above technical solution, the specific method of measuring the actual coordinates of the center point of the calibration object is:

[0033] First, the calibration object is placed within the range scanned by each set of laser radars. Then, the actual coordinates (x0, y0) of the center point of each calibration object are measured by the laser rangefinder. The above design is used to perform square difference calculation with the predicted coordinates of the calibration object center point predicted by the laser radar to obtain the minimum error value.

[0034] In the above technical solution, the stitching process of the stitching point cloud is specifically to obtain the point cloud data information scanned by each laser radar, and stitch the point cloud data information collected by two adjacent laser radars together. The purpose of the above design is that when two adjacent laser radars scan a unified calibration object, the point cloud data information obtained with the laser radar itself as a reference object is adopted, so that the two adjacent laser radars on the left and right of the calibration object collect the point cloud data information of the calibration object with the same reference point, so that the contour information of the calibration object can be stitched out more completely, which is convenient for the subsequent prediction of the center point of the calibration object.

[0035] In the above technical solution, the clustering process of generating the clustered point cloud is as follows:

[0036] Through the clustering algorithm (a density-based clustering algorithm is used in this method, the principle of which is to divide points connected by density into the same cluster and regard points with lower density as noise points; its algorithm steps mainly include two processes: finding core points and merging temporary clusters; it needs to specify two parameters, namely the neighborhood radius ε and the minimum number of neighborhood points Min-Points; the density-based clustering algorithm is only applied to the workpiece scanning method of this method. For actual application conditions, other suitable clustering algorithms or noise filtering algorithms can be used. For ordered or partially ordered point clouds, the nearest neighbor statistical Gaussian noise filtering algorithm, least squares filtering, Wiener filtering and smoothing filtering can be used) to remove outliers in the spliced ​​point cloud that are farther away from the model and the debris scanned by the laser radar to obtain the clustered point cloud. The above design makes the model accuracy higher after clustering denoising.

[0037] In the above technical solution, the point cloud cropping is specifically to manually select the area where the calibration object exists as the area of ​​interest from the clustered point cloud, and then crop it out and put it into a list to generate the cropped point cloud. The above design manually crops the point cloud data information of the calibration object, which can effectively reduce the interference of non-valid outliers, while improving the accuracy of the model, retaining the point cloud needed to predict the center point of the calibration object, reducing the scale of point cloud data, and improving the processing speed of subsequent point clouds and the accuracy of the model.

[0038] In the above technical solution, the specific method for obtaining the predicted coordinates of the center point of the calibration object is:

[0039] By using the point cloud data information of the cropped area of ​​interest, the maximum and minimum values ​​of the horizontal coordinate x and the vertical coordinate y are obtained after sorting. min , x max ,y min ,y max , then we can get the predicted coordinates of the center point of the calibration object as ((x min +x max ) / 2,(y min +y max ) / 2), the purpose of the above design is to obtain the center coordinates of the calibration object predicted after each rotation of the two lidars, so as to facilitate subsequent comparison with the actual center coordinates of the calibration object and find the optimal rotation angle of the left and right lidars.

[0040] In the above technical solution, the calculation process of the error value is:

[0041] The square difference between the predicted coordinates of the center point of the calibration object corresponding to each spliced ​​point cloud and the actual coordinates of the center point of the calibration object is calculated, and the result is used as the error value. The above design can intuitively see the difference between the result obtained by rotating the laser radar angle each time and the actual result, so as to find the optimal time.

[0042] In the above technical solution, the specific process of selecting the calibration correction angle is as follows:

[0043] Compare the error values ​​between the predicted coordinates of the center point of the calibration object corresponding to each stitched point cloud and the actual coordinates of the center point of the calibration object to find the minimum error value. The rotation angle of the left point cloud data corresponding to the minimum error value in the stitched point cloud is the calibration correction angle of the first laser radar on the left side of the calibration object, and the rotation angle of the right point cloud data corresponding to the minimum error value in the stitched point cloud is the calibration correction angle of the second laser radar on the right side of the calibration object. The above design can greatly improve the calibration accuracy of the laser radar.

[0044] Example 2

[0045] A method based on the joint automatic calibration of multiple single-line lidars, such as Figure 2 As shown,

[0046] Measure the actual coordinates of the center point of the calibration object; collect the depth information of the calibration object on the left and the depth information of the calibration object on the right; convert the depth information of the left calibration object into the left point cloud data information of the calibration object, and convert the depth information of the right calibration object into the right point cloud data information of the calibration object; obtain the left point cloud data information of each rotation angle and the right point cloud data information of each rotation angle through the rotation matrix; obtain multiple spliced ​​point clouds, generate clustered point clouds, select the calibration object interest area for cropping, generate a cropped point cloud, and obtain the predicted coordinates of the center point of the calibration object; calculate the error value between the predicted coordinates of the center point of the calibration object corresponding to each spliced ​​point cloud and the actual coordinates of the center point of the calibration object, and use the corresponding left point cloud data rotation angle and right point cloud data rotation angle in the spliced ​​point cloud when the error value is minimum as the calibration correction angle.

[0047] The specific method of joint automatic calibration includes the following steps:

[0048] The laser rangefinder measures the actual coordinates of the center point of the calibration object;

[0049] The first laser radar located on the left side of the calibration object collects the depth information of the calibration object on the left side of the calibration object, and the second laser radar located on the right side of the calibration object collects the depth information of the calibration object on the right side of the calibration object;

[0050] Convert the depth information of the left calibration object into the left point cloud data information of the calibration object, and convert the depth information of the right calibration object into the right point cloud data information of the calibration object;

[0051] The left point cloud data information is rotated N times within the preset rotation angle range by the rotation matrix to obtain N rotation angle left point cloud data information, and the right point cloud data information is rotated M times within the preset rotation angle range by the rotation matrix to obtain M rotation angle right point cloud data information;

[0052] The point cloud data information on the left side of each rotation angle is spliced ​​with the point cloud data information on the right side of each rotation angle to obtain corresponding multiple spliced ​​point clouds; each spliced ​​point cloud is clustered to generate a clustered point cloud; the clustered point cloud is cropped according to the selected calibration object interest area to generate a cropped point cloud, and the predicted coordinates of the calibration object center point are obtained according to the cropped point cloud; the error value between the predicted coordinates of the calibration object center point corresponding to each spliced ​​point cloud and the actual coordinates of the calibration object center point is calculated, and the corresponding left point cloud data rotation angle and right point cloud data rotation angle in the spliced ​​point cloud corresponding to the minimum error value are used as the calibration correction angle.

[0053] Example 3

[0054] A system based on the joint automatic calibration of multiple single-line laser radars, such as Figure 3 、 4 As shown in Figures 5, 6 and 7, it includes:

[0055] Measure the accurate position information of the calibration object. Specifically, place the calibration object in the area scanned by the laser radar, select the coordinate origin, and then use the laser rangefinder to accurately measure the position of the center point of the calibration object, that is, P0 (x0, y0). With an accurate calibration point in the required coordinate system, P0 can be used as a reference point for subsequent calibration work;

[0056] To obtain the point cloud data information of the calibration object, specifically, use a wired network to connect the laser radar and the computer, and then run the main program to start the laser radar scanning to scan and obtain the distance information of the object. The experimental laser radar uses the 3i-T1 series laser radar produced by Sugawa Robotics Co., Ltd. This series of laser radars uses the pulse flight method for distance measurement. The invisible infrared laser emitted by the ranging module is deflected to different angles by a high-speed rotating reflective mirror inside, thereby realizing the scanning measurement of the environmental contour within 270° of the same horizontal plane; the scanning resolution is set to 15Hz, and the distance unit of the laser radar return value is 0.5 mm, so to convert the distance information collected by the laser radar into point cloud information, it is necessary to multiply the collected distance information by 2 and then multiply it by the rotation and translation matrix trans_array=[cos(θ),-sin(θ),t x ; sin(θ), cos(θ), t y; 0,0,1], where θ is the angle when the laser radar scans this point. Subsequent calculations require the calculation of a three-dimensional matrix, so it is necessary to manually expand a dimension to [0, 0, 1], so that the collected data can be converted into point cloud data P0; the point cloud information collected by the first laser radar is as follows Figure 3 As shown in the figure, the point cloud information collected by the second laser radar is as follows: Figure 4 As shown;

[0057] Point cloud rotation, specifically, first determine that the error range of the horizontal angle installation of the laser radar is ±5°, use two nested for loop structures, first make the point cloud information collected by the first laser radar of the calibration object, with the position of the first laser radar as the rotation axis, rotate once every 0.1° in the range of -5° to 5°, each rotation, make the point cloud information collected by the second laser radar of the calibration object with the position of the second laser radar as the rotation axis, rotate once every 0.1° in the range of -5° to 5°, then make the point cloud information collected by the first laser radar and the second laser radar be multiplied by the rotation and translation matrix rotation_matrix_l = [cos(i), -sin(i), t x ; sin(i), cos(i), t y ;0,0,1], rotation_matrix_r=[cos(j),-sin(j),t x ; sin(j), cos(j), t y ; 0,0,1], i, j are the rotation angles of the first laser radar and the second laser radar of the calibration object at this time; t x and t y The translation component is obtained by using these two nested loops. The effect of the horizontal installation angle of the two lidars on the point cloud acquisition within the range of ±5° can be obtained. If more precision is required, each rotation of 0.1 degrees can be changed to 0.05°.

[0058] Point cloud stitching, specifically using the concatenate function in the numpy library, connects the two rotated point cloud data by row; because the data is three-dimensional at this time in order to facilitate the multiplication of the data with the rotation and translation matrix, the delete function in the numpy library needs to be used after stitching to delete the manually expanded dimension, and the result is the coordinates of the object scanned by the required lidar; the purpose of point cloud stitching is that when two adjacent lidars scan a unified calibration object, the point cloud information obtained by using the lidar itself as a reference object is used, so that the first lidar and the second lidar of the calibration object collect the point cloud information of the calibration object with the same reference point, so that the contour information of the calibration object can be stitched out more completely, which is convenient for the subsequent prediction of the center point of the calibration object. Point cloud stitching is as follows Figure 5As shown;

[0059] Point cloud clustering denoising, specifically through the DBSCAN clustering algorithm, which is a density-based clustering algorithm that can identify clusters of any shape and has good robustness to noise points; in Python's scikit-learn library, the DBSCAN class provides a fit_predict method, which combines the fit and predict steps to fit data and predict the cluster labels of data points; its principle is to divide density-connected points into the same cluster and regard points with lower density as noise points; its algorithm steps mainly include finding core points and merging temporary clusters. It needs to specify two parameters, namely the neighborhood radius and the minimum number of neighborhood points; the specific workflow is: cluster each core point (a point with at least min_samples points in the eps neighborhood); add all directly density-reachable points (that is, points in the eps neighborhood of the core point) to the cluster; repeat the above steps until all density-reachable points are added to the cluster; the remaining points are regarded as noise; after processing by the clustering algorithm, the scanned point cloud model is more accurate, and after clustering, Figure 6 As shown;

[0060] Point cloud clipping is specifically done by cutting the point cloud P d Manually select the area where the calibration object exists as the area of ​​interest, set the range of x value and y value of the area of ​​interest, that is, x_range, y_range, and use the command to remove the point cloud outside the area of ​​interest to obtain the point cloud P c (x, y), put them into different lists respectively, so that the coordinates of the center point of the calibration object can be predicted through the point cloud information of the calibration object in different lists. Figure 7 As shown;

[0061] The purpose of clipping the point cloud P0 (x, y) is that when using a lidar to scan an object, all points within the scan range will be generated, and point clouds that do not need to be calculated will be generated. After clipping, irrelevant point clouds are removed, and the point clouds needed to predict the center point of the calibration object are retained, effectively reducing the interference of invalid outliers, while improving the accuracy of the model, reducing the size of the point cloud data, and improving the processing speed of subsequent point clouds and the accuracy of the model.

[0062] Predict the center point of the calibration object. Specifically, the point cloud around the calibration object is obtained in the previous step as the area of ​​interest. Call the min() and max() functions in the numpy library to get the maximum and minimum values ​​of x and y in the area of ​​interest, that is, x min , x max ,y min ,y maxSince the calibration objects generally used are rectangular, the coordinates of the center point of the calibration object measured by the laser radar position can be predicted by these four values, which is (x p ,y p )=((x min +x max ) / 2,(y min +y max ) / 2);

[0063] The purpose of this is to obtain the center coordinates of the calibration object predicted after each rotation of the two lidars, so that we can compare them with the actual center coordinates of the calibration object and find the optimal rotation angle of the first lidar and the second lidar.

[0064] Calculate the error value, specifically (x p ,y p ), and the measured coordinates of the center point of the calibration object (x t ,y t ) to calculate the error, which is loss=(x p -x t ) 2 +(y p -y t ) 2 By comparing the size of each loss, we can intuitively see the deviation between the predicted value and the true value, so that we can find the optimal solution by minimizing the loss in subsequent operations;

[0065] The optimal rotation angle is selected by comparing and sorting the losses obtained in the previous step. That is, the losses obtained when the lidar rotates at different angles are compared with the next loss, and the smaller loss is retained. Then, until both lidars are rotated within the range of ±5°, the minimum error value can be found. Then, the angles at which the first lidar and the second lidar are rotated when the error value is minimum can be obtained. This is the optimal angle in the horizontal position to compensate for errors caused by the installation of the lidar, etc. In this way, there is no need to manually adjust the installation angle repeatedly, thereby greatly improving the calibration accuracy of the lidar and making the measured results closer to the actual results.

[0066] Example 4

[0067] A system based on the joint automatic calibration of multiple single-line laser radars, such as Figure 8 、 9 , 10, it includes:

[0068] Design a graphical interface. Specifically, use QT designer software to design the graphical interface, create and design the UI interface, drag several QPushButtons into it as slots, use clocked() to trigger the signal, and design several buttons to be used for the calibration of the first group of laser radars. Then use Pycharm software to load the dynamic UI interface, use the connect function to fill the designed slot function into the slot, so that the graphical interface can be used to implement specific functions, and then use the information function in the QMessage library to display the results; the designed software interface is as follows Figure 8 As shown, the interface shows the results as Figure 9 As shown;

[0069] Its purpose is to package the required programs and environments so that users can use them directly, providing users with intuitive, easy-to-understand and easy-to-operate visual elements such as buttons, menus, icons, etc., which improves the user experience. Qt is a cross-platform application development framework. GUIs designed with Qt can run on different operating systems, eliminating the need to design separate interfaces for each platform.

[0070] Through the obtained results, the information collected by the laser radar can be rotated to the corresponding angle, so that the data measured by the laser radar can be closer to the real data without manual repeated adjustments, thereby improving the accuracy of the laser radar. The point cloud collected by the adjusted laser radar is as follows: Figure 10 shown.

[0071] Example 5

[0072] A computer program product includes a computer program, characterized in that when the computer program is executed by a processor, the steps of the method described in Example 2 are implemented.

[0073] The contents not described in detail in this specification belong to the prior art known to professional and technical personnel in this field.

Claims

1. A system based on joint automatic calibration of multiple single-line laser radars, characterized in that: It includes: The laser rangefinder is used to measure the actual coordinates of the center point of the calibration object; The first laser radar located on the left side of the calibration object is used to collect the depth information of the calibration object on the left side of the calibration object, and the second laser radar located on the right side of the calibration object is used to collect the depth information of the calibration object on the right side of the calibration object; The point cloud acquisition module is used to convert the depth information of the left calibration object into the left point cloud data information of the calibration object, and convert the depth information of the right calibration object into the right point cloud data information of the calibration object; The point cloud rotation module is used to rotate the left point cloud data information N times within the preset rotation angle range through the rotation matrix to obtain N rotation angle left point cloud data information, and to rotate the right point cloud data information M times within the preset rotation angle range through the rotation matrix to obtain M rotation angle right point cloud data information; The point cloud processing module is used to splice the point cloud data information on the left side of each rotation angle with the point cloud data information on the right side of each rotation angle to obtain corresponding multiple spliced ​​point clouds; cluster each spliced ​​point cloud to generate a clustered point cloud; The clustered point cloud is cropped according to the selected calibration object interest area to generate a cropped point cloud, and the predicted coordinates of the calibration object center point are obtained based on the cropped point cloud; the error value between the predicted coordinates of the calibration object center point corresponding to each spliced ​​point cloud and the actual coordinates of the calibration object center point is calculated, and the corresponding left point cloud data rotation angle and right point cloud data rotation angle in the spliced ​​point cloud corresponding to the minimum error value are used as the calibration correction angle.

2. The system based on joint automatic calibration of multiple single-line laser radars according to claim 1, characterized in that: The specific method of measuring the actual coordinates of the center point of the calibration object includes: First, place the calibration object within the range scanned by each set of laser radars, and then use the laser rangefinder to measure the actual coordinates (x0, y0) of the center point of each calibration object.

3. The system based on joint automatic calibration of multiple single-line laser radars according to claim 1, characterized in that: The stitching process of the stitching point cloud is specifically to obtain the point cloud data information scanned by each laser radar and stitch the point cloud data information collected by two adjacent laser radars together.

4. The system based on joint automatic calibration of multiple single-line laser radars according to claim 1, characterized in that: The clustering process of generating the clustered point cloud is as follows: The clustering algorithm is used to remove outliers outside the model setting range and debris scanned by the lidar in the spliced ​​point cloud to obtain the clustered point cloud.

5. The system based on joint automatic calibration of multiple single-line laser radars according to claim 1, characterized in that: The point cloud clipping is specifically to manually select the area where the calibration object exists as the area of ​​interest from the obtained clustered point cloud, and then clip it out and put it into a list respectively to generate a clipped point cloud.

6. The system based on joint automatic calibration of multiple single-line laser radars according to claim 1, characterized in that: The specific method to obtain the predicted coordinates of the center point of the calibration object is: By using the point cloud data information of the cropped area of ​​interest, the maximum and minimum values ​​of the horizontal coordinate x and the vertical coordinate y are obtained after sorting. min , x max ,y min ,y max , so the predicted coordinates of the center point of the calibration object are ((x min +x max ) / 2,(y min +y max ) / 2).

7. The system based on joint automatic calibration of multiple single-line laser radars according to claim 1, characterized in that: The calculation process of the error value is: The square difference operation is performed between the predicted coordinates of the center point of the calibration object corresponding to each spliced ​​point cloud and the actual coordinates of the center point of the calibration object, and the result is used as the error value.

8. The system based on joint automatic calibration of multiple single-line laser radars according to claim 1, characterized in that: The specific process of selecting the calibration correction angle is as follows: Compare the error values ​​between the predicted coordinates of the center point of the calibration object corresponding to each stitched point cloud and the actual coordinates of the center point of the calibration object to find the minimum error value. The rotation angle of the corresponding left point cloud data in the stitched point cloud corresponding to the minimum error value is the calibration correction angle of the first laser radar on the left side of the calibration object, and the rotation angle of the corresponding right point cloud data in the stitched point cloud corresponding to the minimum error value is the calibration correction angle of the second laser radar on the right side of the calibration object.

9. A method for joint automatic calibration based on multiple single-line laser radars, characterized in that: It follows the steps: The laser rangefinder measures the actual coordinates of the center point of the calibration object; The first laser radar located on the left side of the calibration object collects the depth information of the calibration object on the left side of the calibration object, and the second laser radar located on the right side of the calibration object collects the depth information of the calibration object on the right side of the calibration object; Convert the depth information of the left calibration object into the left point cloud data information of the calibration object, and convert the depth information of the right calibration object into the right point cloud data information of the calibration object; The left point cloud data information is rotated N times within the preset rotation angle range by the rotation matrix to obtain N rotation angle left point cloud data information, and the right point cloud data information is rotated M times within the preset rotation angle range by the rotation matrix to obtain M rotation angle right point cloud data information; The point cloud data information on the left side of each rotation angle is spliced ​​with the point cloud data information on the right side of each rotation angle to obtain corresponding multiple spliced ​​point clouds; each spliced ​​point cloud is clustered to generate a clustered point cloud; The clustered point cloud is cropped according to the selected calibration object interest area to generate a cropped point cloud, and the predicted coordinates of the calibration object center point are obtained based on the cropped point cloud; the error value between the predicted coordinates of the calibration object center point corresponding to each spliced ​​point cloud and the actual coordinates of the calibration object center point is calculated, and the corresponding left point cloud data rotation angle and right point cloud data rotation angle in the spliced ​​point cloud corresponding to the minimum error value are used as the calibration correction angle.

10. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to claim 9 are implemented.

Images