Quick light mixing of multiple color sources for image acquisition using pattern search

Abstract

Illumination is one of the inspection conditions in industrial machine vision and highly related with image quality. Single light source is commonly used and adjusted to acquire fine images during setup. The image acquisition is described by nonlinear and complex equations, and a color mixing source additionally requires multi-dimensional formulation. So, this paper applied a direct, nondifferential, multi-dimensional search method for optimal illumination conditions using pattern search. The pattern search is one of the optimum methods and was modified for this optimal illumination and multiple color sources in machine vision. The pattern search in this paper was discussed about how to organize a probe network for this optimal illumination of image acquisition. The pattern search was composed of probe network of multiple dimensions, probing sharpness, translation, shrinkage, and terminal condition. The proposed method can maximize image sharpness and minimize iterative adjustment in the test results of an RGB mixer, which was more effective than the case of equal step search. The pattern search algorithm for this optimal illumination provides automatic and quick lighting control in image inspection process. The proposed method decreased the iterations under 1% of conventional search, and it is very efficient on time and energy saving.