Robot Taught Path Correction Based On The Corner Detection And Image Matching Algorithm

The main mode of industrial robot’s field application is teaching and playback mode. The position and posture of real-time workpieces are always different from the template workpiece which the robot’s path taught on it. Therefore, object recognition, location and the robot’s taught path correction are important parts of the industrial robot’s automation applications. With the gradually increase of the automaticity, vision sensors are widely used in industrial robots to identification and location the workpieces, guide the robots in industrial environment and so on. Based on the analysis and research in SUSAN corner detection algorithm, SIFT image matching algorithm, camera calibration technology, industrial robot path correction method and some other aspects, this issue proposed a robot taught path correction algorithm which based on the improved SUSAN corner detection and SIFT image matching algorithm.Research on corner detection techniques includes the following contents. Classified a variety of classic algorithms based on the detection principle of each algorithm. Emphatically analyzes the SUSAN corner detection algorithm, experiment shows that there has some poor performance on this algorithm, such as false detection, omitting and so on. To make up the shortfall of SUSAN corner detection algorithm, this paper proposes an improved algorithm based on the neighborhood characteristics of target points. Experiments prove that the improved algorithm has better performance while detect the corners of typical graphics in images than original algorithm, and improved algorithm is much faster than the original algorithm and the classical Harris corner detection algorithm. Proposed a method to determine the corner’s regional direction properties, connect corner points with their direction properties and principle of proximity to create a regional profile, provide a decision basis for workpieces identification in practical application.Analyze the performance of SIFT image matching algorithm in the images with workpieces which take from the industry environment. Proved by experiments that the SIFT algorithm has a good, reliable performance to those images from industry environment, and this method can be used to identify the type of workpieces. Based on the SIFT feature point’s sub-pixel level positioning performance, the difference of position and posture with real-time workpieces and model workpiece can be defined by use of ICP algorithm.Research on the camera model and camera calibrate, calibrate the hand-eye relationship by using an improved SUSAN corner detection algorithm to extract the feature points of checkerboard and using CONTURA G2to measure the feature points’position in World coordinate. Though the research on the robot’s taught path correction method of KUKA robot, this paper propose a simple camera calibration method and Base coordinate system teaching method for KUKA robot system.Finally, with those researches in this paper, a robot’s taught path correction experimental platform which composed of vision sensor, industrial computer, KR16robot was established. Experiments show that the robot taught path correction algorithm which based on the improved SUSAN corner detection and SIFT image matching algorithm can identify and locate the real-time workpieces and correct the robot path.