Research On Laser Radar Target 3D Pose Estimation Algorithm Based On Point Normal Vector

Lidar is a new type of sensor which can obtain the distance image of the target.Through the range profile,point clouds representing the three-dimensional geometric characteristics of the target space can be obtained,which lays a powerful data foundation for attitude estimation Through the three-dimensional attitude information of the target,the specific direction of the target can be obtained,and the attitude estimation is an important link in the automatic recognition of the target.At present,most of the attitude estimation algorithms are still in the simulation range image stage,and the pose estimation of the real scene has not been reported.The three-dimensional pose estimation algorithm based on point normal vector deviates from the actual coordinate value of the target coordinate system.In this paper,a new attitude estimation algorithm(OPDVA)is proposed to estimate the real scene distance image.Firstly,the algorithm firstly corrects the point vector clustering method and the coordinate system of the target coordinate system,and then uses the feature of the large-area approximate plane region in the real scene point cloud to planarly fit the scene plane through the plane model of the random sampling consensus algorithm.The normal vector of the best fitting plane is the coordinate axis of the target coordinate system.Finally,the inverse three-dimensional attitude angle can be solved by the inverse kinematics equation.Using the attitude information obtained,the size of the target is calculated by the rectangular bounding box method.In the experiment,the OPDVA algorithm is verified by simulating the distance image and the real scene distance as two data sources respectively.For the simulated range image,the overall average standard error of the pose estimation is 0.9123°,which is 68.64% of the PDVA algorithm.For the real scene distance image,using the techniques of rotation transformation and resampling,and comparing the rectangular bounding box method with the PDVA method,the accuracy of the OPDVA algorithm for scene pose estimation is obviously better than the other two methods.The overall average standard error is 0.8051°,which drops to 82.76% of the average error of the PDVA algorithm.Occlusion experiments show that the OPDVA algorithm is not sensitive to occlusion.In addition,the measurement error of the building size in the real scene is less than 1 m.The OPDVA algorithm proposed in this paper is more accurate and has certain anti-noise performance than the PDVA algorithm.It can further utilize the spatial attitude relationship between targets to achieve fast target search and detection.