Edge and Corner Extraction Using Particle Swarm Optimisation

Abstract

We apply particle swarm optimisation to the detection of edges and corners as low level features in noisy images and use these features to recognise simple objects. In this approach, the edges and the corners of an object are detected by a particle swarm optimisation algorithm and then the object is classified based on the number of corners and attributes of the edges by a simple fuzzy rule-based classifier. Several simple geometric objects in different locations, scales, and orientations have been used with a variety of impulse noise levels to assess the system. This system can categorise images containing these simple objects with high noise levels more accurately than an existing swarm-based edge and corner detector.