Reserach On Hybrid Alignment Method For Point Cloud And It's Application

The measurement technology of workpiece position and posture plays a very important role in industrial production,especially with the rapid development of intelligent manufacturing technology,accurate and rapid target pose estimation has become the primary problem to be solved in many applications.Although the appearance of machine vision technology has greatly improved the defects of low flexibility and poor adaptability existing in the traditional way of relying on fixture positioning,the conventional machine vision technology has strong pertinence to products and is constrained by lighting conditions,which is still difficult to meet the increasingly urgent requirements of intelligent production.With the reduction of hardware cost of 3D point cloud acquisition and processing,the research of on-line measurement technology of workpiece pose based on 3D point cloud has been quietly developed.Point cloud is an important data source of 3D machine vision technology,and 3D point cloud processing technology is also an important development direction of machine vision technology.Point cloud alignment method is an important part of point cloud processing technology.The current pose of the workpiece can be estimated by matching and calibrating the point cloud obtained by scanning the workpiece in the standard pose and the current pose respectively.However,the influence of the density of point cloud on the aligning accuracy and data processing speed has a certain conflict.In order to measure the position and pose of workpiece point cloud more efficiently and accurately,a point cloud hybrid alignment method is proposed.The main research contents are as follows:(1)The research status of point cloud alignment technology is investigated and analyzed,and the basic theory of point cloud 3D information extraction and spatial transformation is summarized.Combined with image processing technology,a step-by-step hybrid point cloud alignment method is proposed to obtain the spatial mapping relationship between two sets of point clouds,so as to realize the estimation of the current pose of the workpiece.(2)Aiming at the point cloud of the workpiece with regular geometric features such as plane,we uses the statistical method of clustering analysis to extract the geometric features such as the spatial normal vector,and realizes the estimation of the inclination information of the workpiece in two dimensions,and realizes the first step adjustion of the point cloud on this basis.Then the point cloud is transformed into a point cloud gray image,and image processing technology is used to extract the translation amount in two dimensions and the deflection amount in one dimension to realize the second step adjustion.In addition,the workpiece is constrained by at least one dimension because it usually needs a table top support or tool clamping.Therefore,the position and posture of the workpiece in the space can be detected finally.(3)In the application of automatic welding of high-voltage transmission line tower base,the position and posture information of the workpiece to be welded should be provided to the welding robot for different specifications of products.Using the proposed hybrid point cloud calibration method,a set of measuring target pose information for welding robot trajectory methods is proposed.Experimental results show that the proposed method can meet the requirements of accuracy and speed for tower pose estimation.Compared with the SAC-IA method,the measurement error is reduced by 90% and the time consumption is reduced by 76.32%.(4)The hybrid alignment method has been applied to the flatness online detection of automotive battery terminal plate.Because of the uncertainty of the position and attitude of the product when it arrives at the detection station,the proposed method is used to measure the position and attitude of the product,and then the point cloud data is extracted from the region of interest to estimate the flatness.A contrast experiment of flatness estimation based on other point cloud alignment methods and the traditional method of CMM is designed.The experimental results show that the flatness estimation results of the three methods are generally consistent,but the time consumption of the proposed hybrid calibration method is reduced by at least 62%.