| Summary: | 碩士 === 大葉大學 === 機電自動化研究所碩士班 === 95 === This thesis develops a pattern writing system of robot manipulators using machine vision. The hardware is based on a personal computer equipped with a PCL-833 encoder card, a PCL-728 D/A card and a color CCD. Borland C++ is used to write a software system. We also present a robust controller to drive the robot manipulator so that a stable tracking is completed. At first, a pattern image is captured by a digital CCD. We then develop image preprocessing techniques including a media filter to filter noises, two-value method to divided the image into a background and a pattern, and a thining method to obtain an on-pixel frame. The labeling and edge-searching methods are employed to obtain the coordinates of the frame. These image coordinates are transferred to space coordinates of the robot manipulator. Through reverse kinematics, we can obtain the desired tracking degrees of the two-axis manipulator. Then, we develop a robust sliding-mode controller to complete a stable tracking for the pattern writing system. Finally, experiments on a practical system confirm effect of the proposed method.
|
|---|
