|
This paper mainly studied the algorithm of the diver’s recognition based on the image ofdiver, it included three parts: the pre-process of the image, the feature extraction of the imageand the classification and recognition of the diver.Firstly, in the section of the pre-process, it mainly studied the methods of the image’snoise reduction. In this paper, it studied basic methods of noise reduction for sonar images,including the mean method, the median method, the morphological smoothing method and thetechnology based on wavelet for images noise reduction. Under different noise conditions, thepaper studied various kinds of method for noise reduction using the images of diver, fish, andthe schools of fishes. After the study, gives the results and the analysis of the simulation forthe image noise reduction.Secondly, in the section of feature extraction, the paper picked up the shape and texturefeatures of the diver and the fish. The shape features included the target’s area, perimeter, andshape parameter, the analysis for texture feature, mainly studied the method of gray levelco-occurrence matrix and gray-gradient co-occurrence matrix, through these methods, thepaper chose the correct characteristics of divers to carry out the next step for diver recognition.In the section of the feature extraction, because of lots of background areas, we need to splitthe target from the background. In the section of the image segmentation, the paper studiedkinds of methods for segmentation, including iteration method, One-dimensional maximumentropy method, and Otsu method.The third section is diver’s recognition; the diver’s recognition under different methodsof segmentation has been studied. Using the method of minimum distance classifier, and sometexture features, the paper classified the diver and the fish successfully.Finally, part of the target recognition algorithms was achieved on DSP, the results of therecognition algorithms’ achieving on DSP has been given.
|