The Research On Method Of Information Fusing Of Mine Recognition

Underwater object recognition has been the main and hard subject in underwater engineering, especially the object such as submarine, torpedo and mine. In the recent fifty years, the technique about detection mines has developed fast, the new technology for detecting bottomed, partially or completely buried mine in deep water with far distance is all-important. In the world, the technique of detection is far less than that of mining. Hence, the development of mine detection is necessary and exigent.This dissertation is a pre-research on information fusion system for mine recognition based on the mature mine recognition and information fusion theory and technique. The contemporary studies in recognition of mine and information fusion are summarized. The pretreatment method of wavelet package, K-L compressor used to reduce feature dimensions, feature fusion classifier and decision layer fusion algorithms based on evidence theory are researched. The recognition software based on its theory is programmed. At last, the dissertation gives the results of simulation of computer and experiment on data obtained from the lake. The result shows that the fusion on feature extraction on multi-domain can effectively improve the recognition rate of the whole, different algorithms of decision layer fusion based on evidence theory improve the results differently, and all of them get the better recognition effects than the signal one.