Urban land-cover classification based on swarm intelligence from high resolution remote sensing imagery

Urban land-cover classification is one of the most challenging problems in pattern analysis and machine intelligence systems in remote sensing. Dense urban environment sensed by very high-resolution (VHR) optical sensors is even more challenging. Occlusions and shadows due to buildings and trees hide some objects in the scene. Despite its simplicity and usefulness, conventional classification methods have failed to have a high classification accuracy in dense urban areas. The objective of this study is to improve the quality of the land-cover classification. We propose using a Particle Swarm Optimization (PSO) based algorithm for the classification of VHR (0.1 to 1 m) remote sensing data over urban areas. This method is to discover classification rules through simulating the social behavior of animals such as the behaviours of bird flocking. The results for aerial images classification show the significance of this method. PSO-based classifier has been applied to the classification of remote sensing data in dense urban district of Kitchener-Waterloo and has achieved high predictive accuracy of 90%.