This thesis describes a design of Industrial Vision System that can be used for recognizing objects in industrial environment. One of the major issue of the industrial vision system is lengthy processing time due to software implementation and bulk data for image coding. To reduce the processing time, the image data part of software implementation is replaced by hardware implementation so as to compress the image data.
A modified chain coding algorithm is proposed for hardware implementation The algorithm is based on the state of neighbor-value of boundary pixel and tracking mode. In hardware for the algorithm, called "Chain coder", features are extracted from chain code lists, and based on positions of feature space, features of the objects to learn are calculated and registered on a data base.
The designed vision system shows that the chain coding time is 20 times faster than a typical software based vision system, so that a vision system for real-time industrial application is feasible.