A New Approach to Underwater Target Recognition

Due to negative effects of underwater imaging environment and the real-time need of underwater task, a new underwater target recognition system is proposed. New combined invariant moments of underwater images are extracted as the system´s recognition features, and the system´s underwater target classifier is based on neural network which improved by Artificial Fish Swarm Algorithm (AFSA). AFSA is capable of attaining global optimum which can make up drawbacks of traditional BP neural network, such as converging slowly and tending to get into the local optimum. The proposed recognition system has been tested using four different kinds of targets images and disturbed images, targets´ affine invariant features are extracted as the inputs of trained neural network and outputs of network are target classification. Experimental results show that the new system is well-clustering and with high classified accuracy.