Research On Target Recognition Based On Genetic Algorithm

Genetic algorithm(GA), as a computational model simulating the biological evolution process of the genetic: selection theory of Darwin,is a whole new global optimization algorithm and is widely used in many fields with its remarkable characteristic of simplicity, commonability, stability, suitability for parallel processing, high eficiency, and practibility. Compared with any other optimization algorithms, the outstanding excellence of GA is the capability of global optimization.Image matching is one key step in the course of image processing and pattern recognization.But, because of the differences of brightness, noise, etc between template image and image, the results by traditional methods of matching is not satisfied. The thesis introduces general concepts of GA and summarizes the productions. On the basis of traditional methods of matching, using chaos or niche technology as one step of GA are presented. The improved methods used for image recognization are successed.The main content of this article is divided into three parts.The first part introduces research history of genetic algorithm and biological background briefly at the beginning, then discuss the basic realization method of Simple GA(SGA), its operaters, its basic characteristic ,its improvements as well as some applications about how to use genetic algorithm to solve problems. In the second part, some relative b asic theories in image matching is first proposed in brief, then introduce the matching methods based on grey and characters particularly. Last, combining the matching methods with SGA, we validate the image matching methods based on SGA.After that, we present two Improved GAs (IGA, including Chaos GA and Niche GA).Using the chaos operating to improve the search, speed, robust and precision of results, a stong capability of local and global searching is proved by the experiments.Because of the capability of global multiple solutions in search space and hidded parallel processing, we present a new method of image recognization based on niching GA (NGA). It prompts the efficiency and robust in multi-object recognition.In fact, all these problems can be generalized as the problem of searching for aglobal optimal solution in a large solution space, which is the classic application field of genetic algorithm. On the major premise of feasibility of this theory, this article based on the practice of forerunners, has done some further research work about the application of genetic algorithm for image matching and pattern-recognition with a satisfactory result.