Infrared Target Recognition Based On Neural Network Research

With the persistent development of world economy, ocean transportation has become increasingly important. But from the development of navigation technology, we can see that, although IMO and related organizations have set down a series of ship avoiding rules , forced ships to equip necessary navigation aid devices, implemented vessel traffic service system (VTS) and the ship reporting rules, ocean accident happened frequently just the same. Infrared imaging device can detect target day and night passively, and has strong anti-jamming ability. Therefore, making the device an efficient supplement to observation equipments like radar and ARPA can greatly improve the navigation security, and it is obvious that, research on infrared target recognition is of great importance to the future shipping undertaking.Infrared target recognition mainly includes two sides: target detection and tracking. After analyzing its principle and methods, the dissertation applies the neural network (NN) technology which develops rapidly in recent years in infrared target recognition, gives full scope to the intelligence advantage of the NN and has received a satisfying result.The dissertation first analyses the shipping development at home and broad pointing out that the research on infrared target recognition is necessary to develop ocean transportation business and improve ship's navigation security, and puts forward the notion of recognizing the target using NN technology according to the research actuality in the field both here and abroad.Secondly, it introduces the development and extensive applicative fields of NN and describes its basic architecture including model , transfer function , the connection manner and learning style which lay a fundation for the design of NN in the following two passages when we use the NN to detect and track the target.Thirdly, it clarifies the principle and algorithm of target detection and tracking, uses the RBF network and NARX network to realize the target detection in a single image, and uses BP network to assist Kalman filter to realize the target tracking. And the experiment results prove that NN is feasible and effective in target...