Improvement of Localization of Vehicle Inspection Sticker with Projective Transformation and Constraints

Abstract

We propose a dense matching method with a series of constraints under genetic algorithm (GA) [1] frame-work for localization of the vehicle inspection sticker (VIS). In this paper, we obtain massive candidate regions by projective transformation, and measure the fitness of candidate region according to the VIS pixel value distribution based constraints. Then, the problem becomes to search an optimum region in candidate regions with the highest fitness function value. Specially, we search the optimum region by GA. The experiment results show that our method can robustly localize the VIS under obvious environmental changes.