The Research Of DIP Automatic Assembly Technology Based On Visual Robot

With the development of robotics technology, the manual work is replaced by industrial robot in traditional industrial production. Especially the machine vision technology which is applied to the robot to make it detect the outer target can bring the robot the function of vision. With the visual identification of the target, the trajectory of robot can be calculated by the feature information of target. In this dissertation, the camera calibration of machine vision, image pre-processing and feature information detection technologies are combined to research the application of visual robot in the DIP automatic assembly system. The dissertation is divided into three main parts.1) The research of camera calibration of visual robot, the transition matrix among the different coordinates of the image processing and serval main camera calibration technology theories are analyzed detailedly, especially the Zhang Zhengyou camera calibration method. The inner and outer parameters of camera are calculated by the Zhang Zhengyou method, based on the compensation of distortion during the calculation, the results is more precise and the pixel error reduces 80%.2) The research of image pre-processing technology of the DIP assembly system, because of the original component images are collected in an industrial environment, the original images have different kinds of noisy information, the mean filtering algorithm and Gaussian filtering algorithm are combined to filter out the noise and the result images have higher SNR value. The OSTU algorithm is used to get the segmentation result of the component images from the background images. Finally the Canny edge detection algorithm is used to calculate the edge information of component images.3) The location of the DIP assembly components, the automatic assembly of the DIP component and the multi-holes PCB component are the main components, so there is a need to detect the assembly components and locate their positions. As to the DIP component, the frequrently-used corner detection algorithms have been used to detect the corner information of the DIP component, however the corner information is too complicated to locate the positions of DIP pins accurately. The concave-convex detection algorithm is brought to detect the concave points of the pins and the costd time is just 1.8% of the normal corner algorithms cost. As to the detection of multi-holes PCB component, one improved random Hough transform detection method with the compression of parameter space is used to detect the holes imformatin of the multi-holes PCB component, the costed time by this method is reduced to 5% of the costed time of normal random Hough transform.Based on the detection results of the DIP component and PCB component, the image position information can be converted to the robot coordinate to locate their positions, the location error is 0.02 mm and the location has high accuracy. Based on the location, the robot trajectory can be calculated to grab the components and assemble them automatically. The visual robot can be used to replace the manual work in the DIP assembly production to improve production efficiency greatly.