Parameter optimisation of an image processing system using evolutionary algorithms

Abstract

The automatic surface inspection, and thus the determination of quality features, is' favourably done by image processing. In the past, numerous methods to improve the performance of image processing systems have been discussed and tested. Unfortunately, new image processing tools and techniques often increase the number of parameters to be adjusted. In order to use the potentiality offered by these techniques, fine-tuned settings of system parameters need to be provided for each image processing task, thus for each object to inspect. Furthermore, in industry the possibility of adjustment is restricted by the lack of image processing knowledge of the operators. In this paper, we propose a method to automatically aptimise the parameters of a machine vision system for surface inspection by using specific Evolutionary Algorithms (FAs). Especially, we pay attention onto the following items: first, the optimisation of the systems parameters in order to increase the efficiency, and second, the automatic adjustment of these parameters to simplify the using of the system. The specifications for using Evolutionary Algorithms to optimise the parameters of an image processing system entail modifications of the standard EAs. Two modifications are proposed in this paper: selection switch and σ-comparison.