Research And Application Of Feature Extraction For Ocean Images

Metric rectification for a perspective image of ocean surface plane is very important for ocean observation. Once the imaged circular points of a projective plane are identified, then projective distortion may be rectified up to a similarity. But it is a challenge task to identify the imaged circular points of ocean surface plane due to the effects of ocean waves. In this paper, a circular floating object on sea is used to identify the metric rectification matrix of ocean surface plane. Experiments show that the approach can effectively eliminate the effects of wave fluctuation, and perform metric rectification for the image of ocean surface plane.Feature extraction is the premise of metric matrix calculation with regard to the metric rectification method on sea image. At the same time, it is also a classic problem of computer vision, and plays an important role in the field of image processing, machine vision, pattern recognition and so on. The accuracy of image feature extraction impacts directly on the image correction results. What we discuss in this paper is the contour feature extracting of horizon and a circular floating object on sea. As the projection of circular floating object in image plane is ellipse, this paper focuses on the geometrical features extraction of line and elliptic curve.As to horizon extraction, a method called the fusion of color models is been proposed in the paper. This method is based on the mathematical selection model of Markowitz. It combines nine color channels with the optimal set of weights, which converts a color image to a gray scale image, enlarging the difference between the gray value of sky and sea. At last, the horizon is extracted from the intersection line between the sky and sea in the gray image. Image experiments based on sea images photographed in sea ice platform and in different time a day demonstrate the feasibility of the proposed approach.As to the extraction of elliptic curves in sea images, the paper first detects the edge features with the color space fusion method, followed by direct least-squares method to fit the edge features. Because of the relatively complex sea environment, two commonly used ellipse fitting methods are deeply discussed in this paper, which are the method of voting and optimal fitting. By comparing the two we finally find a suitable ellipse fitting method for the environment of sea. Finally, the application of above methods to the images photographed in the progress of jacket launching and the real image rectified experiment tell the feasibility and practicality of the algorithms mentioned in this paper.