Wheel Hub Location Method Based On Image Matching Technique

In the automated processing of visually intelligent wheel hub machining, computers need to analyze the real work images to guide industrial robot to grasp and move wheel hub in time. The wheel hub is casting work piece, after roughing process, the side of wheel hub will remain a lot of casting lines and the surface of the hub is very rough. During practical wheel hub finishing process, wheel hub should be recognized from actual obtained image firstly, then its gas nozzle should be locate so that the grasping posture of the robot could de ensured. Therefore, it is very important to precisely locate the wheel hub and its gas nozzle.The main contributions of this thesis include the following:(1) A template matching based wheel hub location method is proposed. First, locate the wheel hub roughly by minimum cross-entropy. Second, locate the center of the wheel hub and calculate the radius of the circle by Hough transformation. Third, Frieze-Expansion (FE) transformation can expanse the circle into rectangle group. Then we use gas nozzle template to match the location of gas nozzle in rectangle group. Lastly, calculate the position of gas nozzle in original image.(2) A feature matching based wheel hub location method is proposed. The SIFT feature points information need to be extracted and saved offline, then the circle center and air cock position of template image should be marked in advance. In real circumstances of wheel hub location,, we should extract SIFT feature points from to be detected image, then use BBF searching algorithm search for the feature points to match with wheel hub template. Followed by Random Sample Consensus algorithm, which can get rid of the error points and calculate the space mapping relationship between template and the image to be detected. Finally, we can use space mapping relationship to locate the circle center and gas nozzle position.This thesis take a template image under good condition, and use other32images which are in rough condition to test two methods under rough condition. The simulation results show:(1) On idealized conditions, template matching method can locate the wheel hub and gas nozzle precisely. However, this method is very sensitive to the change of the image contrast and grayscale.(2) Feature matching method based on SIFT can overcome the problem of lighting, translation, rotation and change of scale. Hence, this method can adapt to different situation easily.