Self-adaptive model of texture-based target location for intelligent transportation system applications

In this paper, three shortages: full-enhancement mechanism, invariable texture extraction mechanism, and lack of robust enhancement methods, that are always neglected in some tasks based on texture-based target location for Intelligent Transportation System applications (ITS) are introduced. To cover these shortages, a self-adaptive model for texture-based target location is established. It is mainly composed of self-adaptive enhancement mechanism, self-adaptive texture extraction mechanism, and graying enhancement. Also, feedback evaluation strategy for design of texture extraction modes, needed by self-adaptive texture extraction mechanism, is provided. And then, an algorithm of license plate location is designed to evaluate effectiveness of this model. 338 license plate images from natural scenes were applied to test graying enhancement and self-adaptive location. Results of enhancement experiment indicate that graying enhancement is robust in purposiveness, real-time performance, and empirical parameter independence, simultaneously. Further, Results of location experiment demonstrate that not only a sound algorithm of license plate location with high success rate (96.1%) and low average execution time (52 ms) is given but also the effectiveness of the two self-adaptive mechanisms in the proposed model are validated.