ROBOTICS : TWO-DIMENSIONAL IMAGE RECOGNITION WITH ROBOT MANIPULATOR

Abstract

The development of a low-cost vision system is the objective of this project. It was driven by the need for a low-cost and robust vision system in our rapidly changing industries. we have approached and executed this task by a careful selection of algorithms together with a proposed set of new techniques. This is to ensure that the vision system is able to 'fit' into a small micro-computer such as the APPLE II micro-computer. Accomplishing the same task using standard techniques would have required a mini-computer. Software structuring and processing is the key to the vision system. By writing most of the routines in machine language instead of the normal BASIC, processing speeds up to a factor of 200 times have been achieved. Novel techniques to vision processing have been proposed. They include auto-thresholding with the Modified Histogram Approach which gives the added features of versatility and flexibility to adapt to changing lighting conditions. A unique digital filtering technique is also proposed. The dual combination of these two algorithm serves as a powerful pre-processor of image data. Triple tasks of edge detection, boundary connection and object identification are achieved simultaneously by the proposed Circular Contour Following Method. These routines effectively forms the back end portion of the image post-processing. We have tested this system under various conditions and interfaced it to a robot for pick and place operations. It is also suitable for use as a front-end image pre-processor to be linked to a higher-end assembly orientated robot.