关键词: 水下目标提取, OTSU算法, Blob分析, 贝叶斯决策, 数学形态学

Abstract: As it is known that the underwater environment is quite complicated and changeable, as a result, targets and pseudo targets always have a high degree of mixing, and one single segmentation method usually could not abstract ideal target regions. Therefore, this paper proposed a new segmentation method based on Blob analysis and Bayesian design-making. Firstly, the optimistic thresholds were calculated by the improved OTSU algorithm, and then the image was segmented according to this threshold. Through analyzing the connectivity characters, closed contours of regions were achieved. Secondly, the connected regions were described using 7 dimensions of Blob operators and pseudo-target regions were eliminated based on Bayesian decision-making rules. Finally, burrs and disturbances were wiped off through the usage of mathematical morphology operators and ideal target regions were achieved. Through dealing with the images grabbed during the pool experiments using the above method, accuracy and efficiency of the method were verified and the real target regions were acquired.

Key words: underwater targets extraction, OTSU algorithm, Blob analysis, Bayesian design-making, mathematical morphology