Research On Key Techniques Of Visual Robot For Surface Mount Component

With the gradual development of modern manufacturing industry, robots have played a more and more important role in human life. And the emergence of the machine vision makes the robots more intelligent. Lots of information can be got from the outside world through the machine vision system, just like the human eyes, and the interference of the external world can be minimized. The machine vision system can be applied to many worse production environments, because messages can be got from a great distance through it. As a result, the scholars paid more attention to the research of the machine vision, and the research of the machine vision becomes a hot spot in robotics. This thesis put forward an intelligent robot positioning system based on image processing, image identification and camera calibration.This system can be divided into three parts, which includes image processing part, image identification part and camera calibration part.In the image processing part, kinds of image preprocessing methods have been studied on. The truth that the traditional filtering method will makes the image’s edge blurred have been found, through the comparison of the experiment results of median filter, mean filter and Gauss filter. So the Gauss filter is chosen to be the reprocessing method in this system. Some image segmentation methods have been discussed, Ostu was considered to be the best image segmentation method, according to the actual application situation of the robot. Through the comparison of the advantages and disadvantages of Sobel, Roberts, LOG and Canny edge detection operator, Canny is suitable for this system for its high precision.In the image identification part, two kinds of identification method have been studied on. The matching technique based on gray level has many limits and large calculating quantity. While the matching technique based on feature is very steady, efficiency, insensitive to image gray level and image geometric distortion and owes great anti-noise performance. The matching technique based on feature has been selected as the best identification method. After the edge of the target object is obtained, the longest edge of the target object can be got from the use of Hough transform. Then, the deflection angle of the target can be figured out.In the camera calibration part, the pin-hole camera model and the distortion model of the camera lens have been discussed. The calibration method invented by Zhang zhengyou is used in this thesis to obtain the internal and external parameters of the camera. The coordinate transforming relations between the robots coordinates and the image coordinates can be figured out.The positioning accuracy of the system can reach 0.02 mm or less after repeating test, and it satisfied the accuracy requirement of the surface mount component. At present, mount machine is mainly used for SMTs, the robots have not been used for SMTs. But, the mount machine is not only expensive, but also can be used in specific occasions for its lack of versatility. This dissertation integrated the positioning system with the robot control system. Thus, the integrated system not only can reach the desired placement accuracy and speed, but also can be extended to other applications, such as machine parts assembly. The versatility of this system will play a great role in promoting the extensive use of robots.