阅读说明：本技术 一种基于多激光雷达数据融合的方法 (Method based on multi-laser radar data fusion ) 是由 张林灿 董钊志 张婉蒙 宋罡 刘树全 于 2020-05-07 设计创作，主要内容包括：本发明公开了一种基于多激光雷达数据融合的方法,包括如下步骤：读取主激光雷达的原始数据SC1和侧向激光雷达的原始数据SC2；初始化旋转矩阵和平移矩阵；根据步骤二中的旋转矩阵和平移矩阵,变换点云,得到新的点云SC2_；寻找邻近点对,得到重新排列的主雷达点云SC1_；分别去中心得到_SC1_和_SC2_；对应点对矩阵和SVD分解得到新的旋转矩阵和平移矩阵；重复迭代步骤三到步骤六,直到对应点对位置差小于设置的阈值。本发明基于多激光雷达数据融合的方法可以有效地将多个激光雷达的数据充分利用,规避多雷达共同的扫描区域目标物被识别为多个物体,保证对智能驾驶车辆周围环境精准的识别,保证决策算法以及后融合算法的数据有效利用。(The invention discloses a method based on multi-laser radar data fusion, which comprises the following steps: reading the original data SC1 of the main laser radar and the original data SC2 of the lateral laser radar; initializing a rotation matrix and a translation matrix; transforming the point cloud according to the rotation matrix and the translation matrix in the step two to obtain a new point cloud SC2 _; finding a neighboring point pair to obtain a rearranged main radar point cloud SC1 _; respectively removing the center to obtain _ SC1_ and _ SC2 _; decomposing the corresponding point pair matrix and the SVD to obtain a new rotation matrix and a new translation matrix; and repeating the iteration steps from the third step to the sixth step until the position difference of the corresponding point pair is smaller than the set threshold value. The method based on the multi-laser-radar data fusion can effectively and fully utilize the data of a plurality of laser radars, avoids the situation that the target object in the common scanning area of the plurality of radars is recognized as a plurality of objects, ensures accurate recognition of the surrounding environment of the intelligent driving vehicle, and ensures effective utilization of the data of a decision algorithm and a post-fusion algorithm.)

1. A method based on multi-laser radar data fusion is characterized in that: the method comprises the following steps:

the method comprises the following steps: reading the original data SC1 of the main laser radar and the original data SC2 of the lateral laser radar;

step two: initializing a rotation matrix and a translation matrix;

step three: transforming the point cloud according to the rotation matrix and the translation matrix in the step two to obtain a new point cloud SC2 _;

step four: finding a neighboring point pair to obtain a rearranged main radar point cloud SC1 _;

step five: respectively removing the center to obtain _ SC1_ and _ SC2 _;

step six: decomposing the corresponding point pair matrix and the SVD to obtain a new rotation matrix and a new translation matrix;

step seven: and repeating the iteration steps from the third step to the sixth step until the position difference of the corresponding point pair is smaller than the set threshold value.

2. The method based on multi-lidar data fusion of claim 1, wherein: in the second step, the method is realized through a conversion matrix, the conversion matrix is defined as H, and the H is decomposed into a rotation matrix T and a translation vector M

Wherein the content of the first and second substances,M＝[x₀y₀z₀]^T，a₃₀＝a₃₁＝a₃₂＝0，a₃₃＝1；

converting coordinates of the point P, Q in different coordinate systems once, wherein P is TQ + M; for the above transformation, parameter estimation is performed.

3. The method of claim 2, wherein the set of points K, L for a set of coincident regions are obtained first, and an objective function is establishedk_iAnd l_iThe optimal solution of min f (T, M) is solved for the point coordinates in the point sets K, L, respectively.

The technical field is as follows:

the invention relates to a method based on multi-laser radar data fusion, and belongs to the technical field of intelligent driving of electric vehicles.

Background art:

the Multi-sensor Information Fusion (MSIF) is an Information processing process which is carried out by utilizing computer technology to automatically analyze and synthesize Information and data from multiple sensors or multiple sources under a certain criterion so as to complete needed decision and estimation. The fusion is divided into primary data pre-fusion and target data post-fusion according to the realization principle; fusing bottom data of the sensor to obtain primary data pre-fusion; later stage identification results obtained by the sensors are utilized, namely, each sensor independently generates target data, and then the main processor fuses the characteristic data to realize a perception task called later stage target data fusion; most of the current fusion modes are late-stage target data fusion; sensing or positioning tasks such as obstacle detection, lane line detection, semantic segmentation and tracking, vehicle self-positioning and the like are completed by the sensors, and then confidence degrees are added for fusion. The processing algorithm of the late-stage target data fusion mode is relatively simple and can be performed in a modularization mode, but the effect precision is not as good as that of the original data fusion.

The invention content is as follows:

the invention provides a method based on multi-laser radar data fusion for solving the problems in the prior art, which adopts a 32-line main laser radar and two 16-line blind-sweeping lateral laser radars to realize the fusion of Point sets of the original data of the Point cloud overlapping part based on an Iterative Closest Point (ICP) algorithm, thereby realizing the high-precision identification of a target object.

The invention adopts the following technical scheme: a method based on multi-laser radar data fusion comprises the following steps:

the method comprises the following steps: reading the original data SC1 of the main laser radar and the original data SC2 of the lateral laser radar;

step two: initializing a rotation matrix and a translation matrix;

step three: transforming the point cloud according to the rotation matrix and the translation matrix in the step two to obtain a new point cloud SC2 _;

step four: finding a neighboring point pair to obtain a rearranged main radar point cloud SC1 _;

step five: respectively removing the center to obtain _ SC1_ and _ SC2 _;

step six: decomposing the corresponding point pair matrix and the SVD to obtain a new rotation matrix and a new translation matrix;

step seven: and repeating the iteration steps from the third step to the sixth step until the position difference of the corresponding point pair is smaller than the set threshold value.

Further, in the second step, the second step is implemented by a transformation matrix, the transformation matrix is defined as H, and H is decomposed into a rotation matrix T and a translation vector M

Wherein the content of the first and second substances,M＝[x₀y₀z₀]^T，a₃₀＝a₃₁＝a₃₂＝0，a₃₃＝1；

converting coordinates of the point P, Q in different coordinate systems once, wherein P is TQ + M; for the above transformation, parameter estimation is performed.

Further, firstly, a set of point sets K, L of the coincident regions is obtained, and an objective function is establishedk_iAnd l_iThe optimal solution of min f (T, M) is solved for the point coordinates in the point sets K, L, respectively.

The invention has the following beneficial effects: the method based on the multi-laser-radar data fusion can effectively and fully utilize the data of a plurality of laser radars, avoids the situation that the target object in the common scanning area of the plurality of radars is recognized as a plurality of objects, ensures accurate recognition of the surrounding environment of the intelligent driving vehicle, and ensures effective utilization of the data of a decision algorithm and a post-fusion algorithm.

Description of the drawings:

FIG. 1 is a flow chart of a method of the invention based on multi-lidar data fusion.

The specific implementation mode is as follows:

the invention will be further described with reference to the accompanying drawings.

The invention relates to a method based on multi-laser radar data fusion, which comprises the following steps:

the method comprises the following steps: reading the original data SC1 of the main laser radar and the original data SC2 of the lateral laser radar;

step two: initializing a rotation matrix and a translation matrix;

step three: transforming the point cloud according to the rotation matrix and the translation matrix in the step two to obtain a new point cloud SC2 _;

step four: finding a neighboring point pair to obtain a rearranged main radar point cloud SC1 _;

step five: respectively removing the center to obtain _ SC1_ and _ SC2 _;

step six: decomposing the corresponding point pair matrix and the SVD to obtain a new rotation matrix and a new translation matrix;

step seven: and repeating the iteration steps from the third step to the sixth step until the position difference of the corresponding point pair is smaller than the set threshold value.

Among them, the basic principle of the ICP algorithm: in the process of point cloud matching data fusion, the ultimate goal is to convert a plurality of groups of point cloud information under different coordinate systems into a unified coordinate system through one rotation and translation. The above process can be implemented by a transformation matrix, which is defined as H, which can be decomposed into a rotation matrix T plus a translation vector M

Wherein the content of the first and second substances,M＝[x₀y₀z₀]^T，a₃₀＝a₃₁＝a₃₂＝0，a₃₃＝1；

firstly, obtaining point set K, L of a group of overlapped regions, and establishing an objective functionk_iAnd l_iThe point coordinates in the point sets K and L respectively, and the problem is converted into an optimal solution problem for solving min f (T, M).

According to a certain constraint condition, finding out the nearest neighbor point (k)_i,l_i) And solving optimal matching parameters T and M to minimize the error and minimize an error function f (T, M). And (3) solving a least square error by the least square method (solving the variance), if the least square error is smaller than a set value, (or the iteration number reaches an upper limit, or the least square error is not changed in a small range after each iteration), finishing the calculation, otherwise, continuing the iteration.

The method based on the multi-laser-radar data fusion can effectively and fully utilize the data of a plurality of laser radars, avoids the situation that the target object in the common scanning area of the plurality of radars is recognized as a plurality of objects, ensures accurate recognition of the surrounding environment of the intelligent driving vehicle, and ensures effective utilization of the data of a decision algorithm and a post-fusion algorithm.

The foregoing is only a preferred embodiment of this invention and it should be noted that modifications can be made by those skilled in the art without departing from the principle of the invention and these modifications should also be considered as the protection scope of the invention.