Automated design of the field-of-view, illumination, and image pre-processing parameters of an image recognition system

Abstract

Machine vision is playing an increasingly important role in the industrial field, and the automated design of image recognition systems has been the subject of intense research. In this study, we proposed a system that is capable of automatically designing the field-of-view of an image recognition system, based on the relationship between the camera and the target objects, the illumination conditions, and the image preprocessing parameters. We reformulated the design problem as an optimization problem, and used a multi-start nearest neighbor search method to solve it. Two evaluation experiments were conducted, with different distances between the target objects. The results demonstrated that the system was able to choose an appropriate field-of-view, illumination conditions, and image pre-processing parameters, taking account of the distance between target objects and the required accuracy of recognition.