Research on recognition method of UUV vision based on PSO-BP network

Abstract

During unmanned underwater vehicle (UUV) recovery process, the vision sensor needs to extract and recognize all light sources from guided images. Firstly, Aiming at restoring underwater image without exact depth map, a segmented-linear-mapping approach based on scattering model is put forward. Restoring results indicate that the approach is capable of highlighting subtle details and improving image quality effectively. Then the light source edge and its invariant moments are extracted by Canny edge detection algorithm combined with improved Snake model. At last, a kind of Back Propagation network based on particle swarm optimization algorithm (PSOBP) for UUV recognition is designed, which demonstrates higher recognition rate than traditional BP network.