A Research In Patter Recognition Of Underwater Object Based On Image Processing

The subject stems from the project of "information processing and understanding technology of underwater optical vision" in military intelligent underwater robot technology. It is important to smoothly perform the tasks of detecting the underwater objects and getting their position in the complex condition of the underwater.The ATR(automatic target recognition) includes image collection, image processing , feature extraction and pattern recognition. Considering the real task environment of AUV (Autonomous Underwater Vehicle) the work of thesis is mainly concentrated on the edge detection, feature extraction and recognition of the underwater images.First, the paper has done the research on the edge detection. A new rapid edge detection arithmetic base on the classic Kirsch arithmetic operators is used in the thesis, which improves the speed and effect of edge detection.In the next place, the eigenvectors with better clustering effect are got through the feature extraction and selection. Considering the influence of scale factor on moment invariant features in discrete, six moment invariants with scale, translating and rotating invariance are proposed, which are the six feature vectors.Finally, the in-depth research about pattern recognition is carried out. Both the fuzzy clustering recognition and BP neural net method are applied to recognize underwater object. The two methods are tested with six feature vectors. Theoretical and experimental results show that both two ways extract different features of different patterns, the two methods can distinguish different patterns...