Research On Edge Detection Using Ant Colony Optimization

Edge was a pixel set like abrupt cliffs between darker and brighter pixels,which is step or roof edge in a image.Therefore,edge detection is a method that each pixel is calculated by the gradient.However,those methods ignore the continuity of edge when detecting edge point.Ant Colony Algorithm(ACO)is a optimization algorithm inspired by the natural collective behavior of ant species.The major collective behavior is the foraging behavior that guides ants on short paths to their food sources,since ants can deposit pheromone on the ground in order to mark some favorable path that should be followed by other members of the colony.Inspired by this,it is possible to detect more consecutive edge by the dependency of edge point using the ACO.Image edge detection using ACO algorithms is usually built on fixed size populations,however some studies indicate that varying the population size can increase the adaptability of these systems and their capability to react to changing environments.To this end,in this paper we present an extended model of an artificial ant colony system with varying population size designed to evolve on digital image habitats.First of all,the pheromone fields and inspired fields are defined,introducing the search direction weight-inspired operator;Secondly,the relation between the ant decision of strategy and heuristic information is built to guide the ant tests and marks points on the edge position;Next,the mechanisms of aging and reproduction are established to reinforce the convergence of ant colonies and to self-regulated varying the population size.Finally,the contours are extracted based on the location of pixel pheromone.To test the proposal approach,the experiment is carried out on natural images,license plate image and high-resolution remote sensing images.The testing results show that the proposed approach can not only determine the number of ant colonies,but also detect edge well.Furthermore,the results of quantitative evaluation and qualitative evaluation show that the proposed approach works well and is very promising.