Target Recognition Algorithm Based on BP Networks and Invariant Moments

On the basis of multi-sensor fusion algorithm, a target recognition algorithm based on Back Propagation (BP) neural networks and invariant moments was proposed. Invariant moment takes advantage of overall information of the targets. It has good differentiating effect and high identification technique. On the other hand, BP neural networks not only have the adaptive learning ability, but also are insensitive to imperfection of input mode. Therefore, it has proper classification and extensibility. It is effective for the algorithm based on BP neural networks and invariant moments that decrease the adverse impacts for the images, which are always subject to the changes of imaging distance, direction and position. Simulation results show that the algorithm has strong recognition capability for surface targets from infrared image sensors.