The Research Of Spatial Position And Posture Measurement For Stack Box Workpiece Based On Monocular Vision

The current industrial robot is in the stage of rapid development.Due to the rapid development of computer vision and sensor technology,the use of machine vision for target identification and positioning is an important research hotspot.In the procedure of industrial manufacture automatization,one operation that a robot often needs to complete is the capture of an artifact.To obtain the spatial position and posture of the workpiece is the precondition of the capture opetation.For most of the box workpiece to be processed,workers usually stack workpieces directly on the platform.The stack of workpieces will result int the adhesion of the image,at the same time,the problem of the measurement of the plane position is upgraded to the problem of space pose solution.In order to solve the problem of such image adhesion and spatial position measurement,this paper presents an algorithm for the spatial position measurement of stacked box workpieces based on monocular vision.The feature points spatial coordinate measurement model is established by combining the workpiece geometric constraints as well as the unique solution principle of the P4P problem and spatial vector mixed quadrature.Trial-and-error method is used to realize the exact matching of the two sides of the feature plane with the size of the workpiece.The paper has also analyzed the problem of misjudgment caused by the box workpiece stack,and optimized the trial-and-error method.The spatial coordinates in the camera coordinate system are calculated according to the feature point of the workpiecs by using the feature points spatial coordinate measurement model,and then the spatial position of the feature plane is derived.Finally,the spatial position of the box workpiece is derived according to the position relationship between the target plane and the feature plane.The extraction of feature point two-dimensional plane coordinates is the starting point and key point of all the theoretical derivation.This paper has analyzed the disadvantages of directly extract--ing the corner points of the workpiece,and coordinate of intersection point of the edge fitting line is proposed as the plane coordinate of the feature point.The conventional edge of the workpiece is extracted by using Sobel operator and Canny operator,and the adhesive edge of the workpiece is extraced by using SVM segmentation.All edges of the workpiece are solved by edge synthesis and the Hough transform is used to fit the edge line.Finally,the two-dimensional plane coordinates of feature point are solved.The experimental part has given the specific results of the camera calibration.In this paper,a verification experiment of spatial position algorithm is designed.The adjacent side length and spatial position of the feature plane are calculated when the workpiece in coplanar position,collinear position and general position.Compared with the theoretical value,the average error of side length is 1.02%.The experiment verified that the proposed algorithm is reliable.