Research On Localization And Detection Algorithm For Automotive Door Based On Vision

With the rapidly development of China’s industrial economy,in recent years,the automotive market presents the trend of a steady growth.Therefore,the automation and intelligent of the automotive industry was put forward higher requirements.Realizing the visual positioning of the automotive door in the productive process is an important step to improve the automation level and the level of artificial intelligence.In order to realize the automation of grasping the door of the automobile production line,it is very important to study the algorithm of the localization and detection for automotive door.With the combination of machine vision technology and robot technology,the key technology to solve is the calibration of vision system,which include the coordinates transformation of the joints of the robot and the calibration of relative position between camera and robot.This is also the key problem of converting pixel coordinates into the base coordinates of the robot.In this paper,a fast calibration method for vision system that is new and quite easy is raised.Firstly,the method sets the target coordinate system at a certain position of the calibration plate,and sets the workpiece coordinate system of the robot to coincide with the imaginary target coordinate system.Then the calibration of the workpiece coordinate system of the robot is completed by the three-point method.Finally,the pixel coordinates of the target points are expressed by the base coordinates of the robot.Without any complicated calibration procedures for the internal and external parameters of the camera and the traditional matrix derivation procedure for Eye-To-Hand calibration,this method realizes the acquisition for the 3D information of the plane part by the single-eye camera.Image processing is the key step for the recognition of automotive door,including image preprocessing,image segmentation and image matching,in which the image matching is the most important step.This paper mainly studies the current mainstream point-based feature matching and template-based matching method,and proves the feasibility of pattern-based template matching.In this method,the edge of the image is extracted by the Canny edge detection,and and uses the Freeman chain code to express the extracted target edge.Finally,the matching of contours is realized by Hu moments.Combining with the demand of production,the robot need to get the location and the rotation angle of the door.In this paper,the position of the door is obtained by using the center of gravity of the contour,and the rotation angle of the workpiece is obtained by using the minimum bounding rectangle of the contour.The experiment proves that the method can locate the door accurately.After completing the study of visual system calibration,coordinate transformation and door positioning,the paper builds the software system on the VS platform using C #,and completes the related tests on the system,which verifies the reliability and practicability of the system.