Image Processing Algorithms Used in Robotics Applications

Abstract

The paper presents the results obtained in implementing image-processing algorithms (edge detection, contour closing, skeletonizing). Several classical as well as two original algorithms are presented, together with sample images from the implementation. Image processing for robotic applications requests fast computation times. We investigate the possibility to improve the processing time of a classical skeletonizing algorithm (Nacache and Shingal method) using a previously trained neural network. Two approaches are considered: a two layers neural network and a cluster of 12 single layer neural network. The processing times obtained are presented and commented.