Research-based Underwater Acoustic Images Underwater Target Recognition

Underwater target recognition system is central issue because of military confrontation,to defend the coastal defense and civilian resources, energy development, such as submarinedetection both the enormous benefits in recent years. Divers cost lower funds, and greaterfeatures,at this point, Divers gradually become a tool to use by the navies. It is of greatsignificance for my thesis to research divers underwater target recognition.This paper introduces the complete structure of the general target recognition processfirst, studies the difficulties caused by the system effect is unique in the underwater acousticimage processing and underwater acoustic conditions, And then research the underwatertarget recognition system.Pre-processing:The main purpose of Image pre-processing of underwater acoustic imagepre-processing is denoising. In this paper, a general image processing method is using todenoise algorithms to the study of underwater acoustic images, after the principle ofverification and simulation results to compare their respective advantages and disadvantages,the paper give several universal algorithm for general underwater acoustic image theconclusion which the effect of better wavelet hard thresholding, soft thresholding in-depthstudy, and gives a innovative algorithm, the experiments show that better results inunderwater acoustic images.Image feature extraction: In this paper, divers images and stock image to target image,the image segmentation is also the pretreatment part of the essential link. Iterative thresholdOtsu given image segmentation, the contrast drawn iterative threshold segmentation is moresuitable for this water supply sound image morphological processing, so that it can be usedfor further feature extraction.The feature extraction part is key effects of the entire underwatertarget identification system. This paper analyzes the characteristics of the water to the soundpicture, selected based on the characteristics of the classic shape features extraction method ismore applicable to divers images, and feature extraction amount done a normalized, for thefollow-up process to improve efficiency.target recognition: neural network is a hot topic in pattern recognition. Many advantagesto making the most of today’s image recognition system are selected for target recognition.The selected BP neural network simulation of the extracted features based on the classic shapefeatures, the results show very good results.