Swarm Intelligence Algorithms In Image Matching Applications

Image matching is a procedure which determines the template image is in the big search image or not, and further to find out the best matched location by a certain algorithm. It is one of the key steps in image analysis and computer vision. Image matching plays a very important role in robot vision, target tracking, automatic navigation, motion analysis, character recognition and so on. In recent decades, domestic and foreign researchers have done a lot of work in image matching, and have proposed a large number of image matching methods. However, these methods are not perfect yet in matching accuracy, speed, versatility and robustness.In this paper, we focus on two new swarm intelligence optimization algorithms, Artificial Bee Colony Algorithm (ABCA) and Artificial Fish Swarm Algorithm (AFSA) and discuss fast and robust image matching methods. The main contributions are as follows:1) Principles, advantages and disadvantages of current image matching methods are discussed and analyzed. Image matching based on swarm intelligence optimization algorithm, especially particle swarm algorithm and genetic algorithm, are described in detail.Commonly, traditional image matching methods adopt complete trial-and-error method to search the best matching. They need to traverse huge amount of points, which must decrease the speed of image matching. Therefore, researchers strive to explore some optimizing algorithms. Genetic Algorithm (GA) and particle swarm optimization (PSO) are two widely used optimized algorithms. At present, either basic or improved algorithms of GA and PSO have been widely used in image matching and have achieved some satisfying fruits.2) A new swarm intelligence optimized algorithm-ABCA is discussed and analyzed first, and then it is introduced to image matching in order to improve the speed of image matching. On the other hand, grey theory is employed to grayscale image matching to analysis the histogram and constructs the fitness function of ABCA, Finally, a fast image matching method based on ABCA and grey relational analysis is suggested.In natural environment, worker bees always can find out multi-honey nectar source quickly and accurately, showing their strong group ability. Similarly, image matching is to locate the best matching point in many possible points. Therefore, it is feasible to use artificial bee colony in image matching and artificial bee colony may have their special swarm intelligence advantages in image matching due to their different actions. Our experimental results showed that the proposed method not only improves the matching speed, matching accuracy, but also has strong noise immunity.3) A new swarm intelligence optimized algorithm - Artificial Fish Swarm Algorithm is discussed, and then a fast image matching method based on wavelet transform and improved artificial fish swarm (WAFSA) is proposed.Artificial fish swarm is behavior-based, intelligent and autonomous. Although the fish don't have complex logical ability and comprehensive judgment like human being, they can find high quality food quickly and accurately by simple individual behavior and special communication.Basic artificial fish swarm algorithm think that artificial fishes are randomly located in a region and freely searching food source in their respective field. Our WAFSA method optimizes the initial population, so that the artificial fishes could be evenly distributed around the searching space, which undoubtedly increases the overall search capability. Moreover, adaptive moving speeds and visions and discrete wavelet transform are used to improve the matching speed. Our experimental results of GASFA showed obvious improvement in matching speed, accuracy and noise immunity.