A laser radar with data acquisition capabilities for high-density point clouds

Abstract

The present invention discloses a laser radar with a data acquisition function of a high-density point cloud, including a laser radar body and a vertically arranged pan-tilt support rod, a pan-tilt rotating component is installed at the top of the pan-tilt support rod, a position sensor is installed on the pan-tilt rotating component, which is used to detect the real-time position of the rotating pan-tilt rotating component, and then detect the rotation angle of the laser radar body, a horizontally arranged rotating arm is installed on the pan-tilt rotating component, and a mounting seat is provided at the outer end of the rotating arm, and the laser radar body is fixed on the mounting seat in a vertical scanning state. In the present invention, the laser radar body is installed vertically to solve the problem of poor vertical angular resolution, the pan-tilt rotating component is rotated horizontally to solve the problem of poor horizontal angular resolution, and the data positioning problem collected by the laser radar body is solved by combining the position sensor, and finally, the high-density point cloud 3D modeling is realized through intelligent algorithms.

Description

Technical Field

[0001] The present invention relates to the field of laser radar technology, and in particular to a laser radar with a high-density point cloud data acquisition function. Background Art

[0002] Most laser radars on the market use horizontal and fixed installation methods to achieve radar detection of objects. The detection distance generally does not exceed 100 meters. In addition, due to the influence of angular resolution, the point cloud density decreases with the increase of detection distance. It is impossible to build a real and effective point cloud 3D model based on the detection results. Especially in applications where the detection distance requires 150 meters to 300 meters, the application products on the market cannot meet the use requirements. Therefore, we provide a laser radar with high-density point cloud data acquisition function. Summary of the invention

[0003] In order to solve the defects of the prior art, the present invention provides a laser radar with a high-density point cloud data acquisition function.

[0004] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0005] The present invention discloses a laser radar with a data acquisition function of a high-density point cloud, comprising a laser radar body and a vertically arranged pan-tilt support rod, a pan-tilt rotating component is installed at the top end of the pan-tilt support rod, a position sensor is installed on the pan-tilt rotating component, and is used to detect the real-time position of the rotating pan-tilt rotating component, and then detect the rotation angle of the laser radar body, a horizontally arranged rotating arm is installed on the pan-tilt rotating component, and a mounting seat is provided at the outer end of the rotating arm, and the laser radar body is fixed on the mounting seat in a vertical scanning state;

[0006] During operation, the gimbal rotating component is used to drive the laser radar body to perform rotational scanning with the gimbal rotating component as the center, and the position sensor is used to collect real-time rotational position information data of the gimbal rotating component; thereby synchronously realizing the collection of radar data information and rotation angle data information detected during the rotation scanning of the laser radar body, and then completing the data collection of high-density point cloud, and finally performing high-density point cloud modeling through intelligent algorithms.

[0007] As a preferred technical solution of the present invention, the intelligent algorithm for high-density point cloud modeling is:

[0008] Step 1: When the laser radar body is not rotating, source point cloud data p0 is collected through the laser radar body;

[0009] Step 2: The source point cloud data p0 is translated and rotated to obtain the target point cloud data q0, thereby obtaining a target point cloud data set Q containing a number of target point cloud data q0;

[0010] Step 3: Perform ICP calculation on the target point cloud data set Q of two adjacent frames to achieve spatial matching of the target point cloud data set Q and establish a new corresponding relationship;

[0011] Step 4: Use a statistical filter to perform inter-frame position filtering on the data obtained in step 3 to remove obvious outliers, thereby achieving the function of downsampling and smoothing the data;

[0012] Step 5, loop through steps 3 and 4, and reduce the amount of computational data by half each time. After multiple algorithm iterations, the data converges to the specified threshold Th to obtain high-density point cloud data M.

[0013] Step 6: Merge the high-density point cloud data M to build a 3D model of the high-density point cloud.

[0014] As a preferred technical solution of the present invention, the method of translating and rotating the source point cloud data p0 to obtain the target point cloud data q0 is to rotate the radar to obtain the point cloud data q0 in step 2, where q0 = R0*p0 = R x (α)*R y (β)*R z (γ)*p0, where R0 represents the rotation matrix; α, β, and γ represent the deflection angles in the x, y, and z directions respectively; R x (α), R y (β), R z (γ) are the rotation matrices in the deflection directions of x, y, and z directions respectively;

[0015] in and

[0016] R0=R x (α)*R y (β)*R z (γ)

[0017] but

[0018] If the pan-tilt rotation angle is θ, the corresponding point cloud data is, q θ =R z (θ)*q0=R z (θ)*R0*P0where

[0019] The gimbal rotates one circle, generating N = 360 / θ point cloud data, qn = R z(nθ)*q0, n=0,1,…,N-1; by cascading these N point cloud data in sequence, we can obtain the target point cloud data set Q={q0,q1,…,qN-1}. If a 16-line radar is used, then Q={Q1,Q2,…,Q16}.

[0020] As a preferred technical solution of the present invention, an ICP calculation method for a target point cloud data set Q of two adjacent frames is to perform ICP operation on adjacent frame planes Q1 and Q2 to obtain plane H1; perform ICP operation on adjacent frame planes Q3 and Q4 to obtain plane H2; perform ICP operation on all adjacent frame planes in sequence, and after one ICP calculation of 16 frames of target point cloud data, 8 frames of point cloud data are obtained, which are represented by H1, ..., H8 respectively.

[0021] As a preferred technical solution of the present invention, the method of using a statistical filter to perform inter-frame position filtering is:

[0022] Step 1: Calculate the average distance d from each point to all K neighboring points.

[0023] Where pi(x i ,y i , z i ), p j (x j ,y j , z j );

[0024] Step 2: Calculate the average value and sample standard deviation of the average distance between all points in the previous step and the K neighboring points;

[0025] Step 3: Compare the distance threshold with the distance of each point in turn. Points exceeding the threshold are marked as outliers and removed.

[0026] Step 4: Calculate the smoothness of each point and sort it, and select features.

[0027] Let the smoothness be C, then S is the number of neighborhood point sets of the point, and r is the element of the neighborhood point set of the point;

[0028] The point cloud data H1, ..., H8 are filtered and smoothed as described above to obtain the corresponding data L1, ..., L8.

[0029] The beneficial effects of the present invention are as follows: the laser radar with the data acquisition function of high-density point cloud, wherein the laser radar body is installed vertically to solve the problem of poor vertical angular resolution, the pan-tilt rotating component is rotated horizontally to solve the problem of poor horizontal angular resolution, and the data positioning problem collected by the laser radar body is solved by combining the position sensor, and finally the high-density point cloud 3D modeling is realized through the intelligent algorithm. The high-density point cloud 3D modeling is carried out through the characteristic intelligent algorithm, and after multiple iterations of the intelligent algorithm, the data converges to the specified threshold to obtain the high-density point cloud data M. Compared with the source point cloud data p0, the amount of point cloud data contained in M ​​is N times that of p0. At the same time, the correlation between the data is enhanced during the iteration process, and the 3D model establishment of the high-density point cloud can be realized. In addition, the present invention adopts the method of vertical installation combined with pan-tilt rotation to effectively solve the problem of laser radar angular resolution for long-distance object recognition. The laser radar detection data is combined with the position sensor data, which greatly improves the point cloud density of long-distance objects, and realizes the 3D model construction function of high point cloud density through a unique intelligent algorithm. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] The accompanying drawings are used to provide a further understanding of the present invention and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention and do not constitute a limitation of the present invention. In the accompanying drawings:

[0031] Figure 1 It is a structural schematic diagram of a laser radar with a high-density point cloud data acquisition function according to the present invention;

[0032] Figure 2 It is a principle block diagram of an intelligent algorithm of a laser radar with a high-density point cloud data acquisition function according to the present invention.

[0033] In the figure: 1. Laser radar body; 2. Pan-tilt support rod; 3. Pan-tilt rotating part; 4. Position sensor; 5. Rotating arm; 6. Mounting base. DETAILED DESCRIPTION

[0034] The preferred embodiments of the present invention are described below in conjunction with the accompanying drawings. It should be understood that the preferred embodiments described herein are only used to illustrate and explain the present invention, and are not used to limit the present invention.

[0035] Example: Figure 1As shown, the present invention is a laser radar with a data acquisition function of a high-density point cloud, comprising a laser radar body 1 and a vertically arranged pan-tilt support rod 2, a pan-tilt rotating component 3 is installed at the top of the pan-tilt support rod 2, a position sensor 4 is installed on the pan-tilt rotating component 3, and is used to detect the real-time position of the rotating pan-tilt rotating component 3, and then detect the rotation angle of the laser radar body 1, a horizontally arranged rotating arm 5 is installed on the pan-tilt rotating component 3, and a mounting seat 6 is provided at the outer end of the rotating arm 5, and the laser radar body 1 is fixed on the mounting seat 6 in a vertical scanning state;

[0036] During operation, the gimbal rotating component 3 is used to drive the laser radar body 1 to perform rotational scanning with the gimbal rotating component 3 as the center, and the position sensor 4 is used to collect real-time rotational position information data of the gimbal rotating component 3; thereby synchronously realizing the collection of radar data information and rotation angle data information detected during the rotation scanning of the laser radar body 1, and then completing the data collection of high-density point cloud, and finally performing high-density point cloud modeling through intelligent algorithms.

[0037] like Figure 2 As shown, the intelligent algorithm for high-density point cloud modeling is:

[0038] Step 1: When the laser radar body is not rotating, source point cloud data p0 is collected through the laser radar body;

[0039] Step 2: The source point cloud data p0 is translated and rotated to obtain the target point cloud data q0, thereby obtaining the target point cloud data set Q containing a plurality of target point cloud data q0; the method for translating and rotating the source point cloud data p0 to obtain the target point cloud data q0 is to rotate the radar to obtain the point cloud data q0 in step 2, wherein q0 = R0*p0 = R x α*R y β*R z γ*p0, where R0 represents the rotation matrix; α, β, and γ represent the deflection angles in the x, y, and z directions respectively; R x (α), R y (β), R z (γ) are the rotation matrices in the deflection directions of x, y, and z directions respectively;

[0040] in and

[0041] R0=R x α*R y β*R z γ

[0042] but

[0043] If the pan-tilt rotation angle is θ, the corresponding point cloud data is, q θ =R z (θ)*q0=R z (θ)*R0*P0

[0044] in

[0045] The gimbal rotates one circle, generating N = 360 / θ point cloud data, qn = R z (nθ)*q0, n=0,1,…,N-1; by cascading these N point cloud data in sequence, we can obtain the target point cloud data set Q={q0,q1,…,qN-1}. If a 16-line radar is used, then Q={Q1,Q2,…,Q16}.

[0046] Step 3, perform ICP calculation on the target point cloud data set Q of two adjacent frames to achieve spatial matching of the target point cloud data set Q and establish a new corresponding relationship; the ICP calculation method for the target point cloud data set Q of two adjacent frames is to perform ICP operation on adjacent frame planes Q1 and Q2 to obtain plane H1; perform ICP operation on adjacent frame planes Q3 and Q4 to obtain plane H2; perform ICP operation on all adjacent frame planes in turn, and after one ICP calculation for 16 frames of target point cloud data, 8 frames of point cloud data are obtained, which are represented by H1,...,H8 respectively.

[0047] Step 4: Use a statistical filter to perform inter-frame position filtering on the data obtained in step 3 to remove obvious outliers, thereby achieving the function of downsampling and smoothing the data; the method of using a statistical filter to perform inter-frame position filtering is:

[0048] Step 1: Calculate the average distance d from each point to all K neighboring points.

[0049] Where pi(x i ,y i , z i ), p j (x j ,y j , z j );

[0050] Step 2: Calculate the average value and sample standard deviation of the average distance between all points in the previous step and the K neighboring points;

[0051] Step 3: Compare the distance threshold with the distance of each point in turn. Points exceeding the threshold are marked as outliers and removed.

[0052] Step 4: Calculate the smoothness of each point and sort it, and select features.

[0053] Let the smoothness be C, then where r i represents a point in the domain of S in the set S;

[0054] The point cloud data H1, ..., H8 are filtered and smoothed as described above to obtain the corresponding data L1, ..., L8.

[0055] Step 5, loop through steps 3 and 4, and reduce the amount of computational data by half each time. After multiple algorithm iterations, the data converges to the specified threshold Th to obtain high-density point cloud data M.

[0056] Step 6: Merge the high-density point cloud data M to build a 3D model of the high-density point cloud.

[0057] The laser radar body 1 is installed vertically to solve the problem of poor vertical angular resolution, and the pan-tilt rotating component 3 rotates horizontally to solve the problem of poor horizontal angular resolution. The data positioning problem collected by the laser radar body 1 is solved by combining with the position sensor. Finally, the high-density point cloud 3D modeling is realized through the intelligent algorithm. The high-density point cloud 3D modeling is carried out through a characteristic intelligent algorithm. After multiple iterations of the intelligent algorithm, the data converges to the specified threshold to obtain high-density point cloud data M. Compared with the source point cloud data p0, the amount of point cloud data contained in M ​​is N times that of p0. At the same time, the correlation between the data is enhanced during the iteration process, and the 3D model of the high-density point cloud can be established. In addition, the present invention adopts a vertical installation combined with pan-tilt rotation to effectively solve the problem of laser radar angular resolution for long-distance object recognition. The laser radar detection data is combined with the position sensor data to greatly improve the point cloud density of long-distance objects, and the 3D model construction function of high point cloud density is realized through a unique intelligent algorithm.

[0058] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A laser radar with high-density point cloud data acquisition function; characterized in that: The invention comprises a laser radar body (1) and a vertically arranged pan-tilt support rod (2), a pan-tilt rotating component (3) being installed at the top end of the pan-tilt support rod (2), a position sensor (4) being installed on the pan-tilt rotating component (3) for detecting the real-time position of the rotating pan-tilt rotating component (3), and further detecting the rotation angle of the laser radar body (1), a horizontally arranged rotating arm (5) being installed on the pan-tilt rotating component (3), and a mounting seat (6) being provided at the outer end of the rotating arm (5), and the laser radar body (1) being fixed on the mounting seat (6) in a vertical scanning state; When working, the cradle head rotating component (3) is used to drive the laser radar body (1) to perform rotational scanning with the cradle head rotating component (3) as the center of the circle, and at the same time, the position sensor (4) is used to collect real-time rotational position information data of the cradle head rotating component (3); thereby synchronously collecting radar data information and rotation angle data information detected during the rotational scanning of the laser radar body (1), thereby completing the data collection of high-density point clouds, and finally performing high-density point cloud modeling through intelligent algorithms. The intelligent algorithm for high-density point cloud modeling is: Step 1: When the laser radar body is not rotating, source point cloud data p0 is collected through the laser radar body; Step 2: The source point cloud data p0 is translated and rotated to obtain the target point cloud data q0, thereby obtaining a target point cloud data set Q containing a number of target point cloud data q0; Step 3: Perform ICP calculation on the target point cloud data set Q of two adjacent frames to achieve spatial matching of the target point cloud data set Q and establish a new corresponding relationship; Step 4: Use a statistical filter to perform inter-frame position filtering on the data obtained in step 3 to remove obvious outliers, thereby achieving the function of downsampling and smoothing the data; Step 5, loop through steps 3 and 4, and reduce the amount of computational data by half each time. After multiple algorithm iterations, the data converges to the specified threshold Th to obtain high-density point cloud data M. Step 6: Merge the high-density point cloud data M to build a 3D model of the high-density point cloud.

2. The laser radar with high-density point cloud data acquisition function according to claim 1, characterized in that: The method of performing translation and rotation on the source point cloud data p0 to obtain the target point cloud data q0 is to rotate the radar to obtain the point cloud data q0 in step 2, wherein q0 = R0*p0 = R x (α)*R y (β)*R z (γ)*p0, where R0 represents the rotation matrix; α, β, and γ represent the deflection angles in the x, y, and z directions respectively; R x (α), R y (β), R z (γ) are the rotation matrices in the deflection directions of x, y, and z directions respectively; in and R0=R x (a)*R y (b)*R z (c) but If the pan-tilt rotation angle is θ, the corresponding point cloud data is, q θ =R z (θ)*q0=R z (θ)*R0*P0 in The gimbal rotates one circle, generating N = 360 / θ point cloud data, qn = R z (nθ)*q0, n=0,1,…,N-1; by cascading these N point cloud data in sequence, we can obtain the target point cloud data set Q={q0,q1,…qN-1}. If a 16-line radar is used, then Q={Q1,Q2,…,Q16}.

3. The laser radar with high-density point cloud data acquisition function according to claim 2, characterized in that: The ICP calculation method for the target point cloud data set Q of two adjacent frames is to perform ICP operation on the adjacent frame planes Q1 and Q2 to obtain plane H1; perform ICP operation on the adjacent frame planes Q3 and Q4 to obtain plane H2; perform ICP operation on all adjacent frame planes in turn, and after one ICP calculation of 16 frames of target point cloud data, 8 frames of point cloud data are obtained, which are represented by H1,...,H8 respectively.

4. The laser radar with high-density point cloud data acquisition function according to claim 3, characterized in that: The method of using a statistical filter to perform inter-frame position filtering is: Step 1: Calculate the average distance d from each point to all K neighboring points. where pi(x i y i ,z i ),p j (x j y j ,z j ): Step 2: Calculate the average value and sample standard deviation of the average distance between all points in the previous step and the K neighboring points; Step 3: Compare the distance threshold with the distance of each point in turn. Points exceeding the threshold are marked as outliers and removed. Step 4: Calculate the smoothness of each point and sort it, and select features. Let the smoothness be C, then S is the number of neighborhood point sets of the point, and r is the element of the neighborhood point set of the point; The point cloud data H1, ..., H8 are filtered and smoothed as described above to obtain the corresponding data L1, ..., L8.

Images