A Design Methodology For Industrial Vision Systems

Abstract

The cost of design, rather than that of target system hardware, represents the principal factor inhibiting the adoption of machine vision systems by manufacturing industry. To reduce design costs to a minimum, a number of software and hardware aids have been developed or are currently being built by the authors. These design aids are as follows: a. An expert system for giving advice about which image acquisition techniques (i.e. lighting/viewing techniques) might be appropriate in a given situation. b. A program to assist in the selection and setup of camera lenses. c. A rich repertoire of image processing procedures, integrated with the Al language Prolog. This combination (called ProVision) provides a facility for experimenting with intelligent image processing techniques and is intended to allow rapid prototyping of algorithms and/or heuristics. d. Fast image processing hardware, capable of implementing commands in the ProVision language. The speed of operation of this equipment is sufficiently high for it to be used, without modification, in many industrial applications. Where this is not possible, even higher execution speed may be achieved by adding extra modules to the processing hardware. In this way, it is possible to trade speed against the cost of the target system hardware. New and faster implementations of a given algorithm/heuristic can usually be achieved with the expenditure of only a small effort. Throughout this article, the emphasis is on designing an industrial vision system in a smooth and effortless manner. In order to illustrate our main thesis that the design of industrial vision systems can be made very much easier through the use of suitable utilities, the article concludes with a discussion of a case study: the dissection of tiny plants using a visually controlled robot.