Research on technologies of underwater feature extraction and target location based on binocular vision

Abstract

An underwater binocular vision system for AUV(Autonomous Underwater Vehicle) to complete autonomous operation task has been developed in this paper. It consists of two parts: feature extraction and target location. In feature extraction, an improved edge detection method has been proposed according to the problem that the noise and false edges existing in the result image by using the traditional Canny edge detection algorithm, this paper has employed morphological operation to remove image noise and increased the extraction of two oblique direction gradient information to obtain the gradient image, then achieved the final edge detection result by using the connect thresholds which came from the Otsu algorithm to connect edges of image; in addition, an improved circle extraction method has been proposed according to the problem that computational complexity and time consuming by using the traditional Hough transform algorithm. In target location, considering the distortion error during the imaging, this paper has added the second order radial distortion into pinhole imaging model to improve the location precision of binocular vision. The validity of these methods is verified by the underwater comparative experiments.