A novel hybrid approach based on a chaotic cloud gravitational search algorithm to complicated image template matching

Template matching is the process of accurately extracting the interesting regions in a source image according to reference templates. In this paper, the gravitational search algorithm (GSA) is employed as a novel search strategy for template matching. However, the basic GSA is easily trapped in a local optimum and has a poor exploitation ability. In this paper, to enhance the optimization performance of GSA, a novel cross-search strategy based on chaotic global search (CGS) and cloud local search (CLS) is incorporated into GSA. The new variant is named chaotic cloud GSA (CCGSA). CGS makes full use of the ergodicity of chaos theory to improve global search ability and to avoid premature convergence. Inspired by the randomness and stable tendency of the normal cloud model, CLS was formed to realize a refined exploitation in the neighborhood of the current best solution; therefore, it can enhance optimization efficiency. Comparative experiments on six composite benchmark functions indicate that CCGSA convergence performance is superior to that of two advanced variants of GSA. Moreover, when applied to template matching, CCGSA performs better than the other selected intelligent optimization algorithms.