| dc.description.abstract | Abstract:
The linear of the earth’’s surface occupies the extremely important part in the research of geosciences. In the past extraction linear of the image usually work done by hand. However, resolution of the image increases day by day, and the quantity of the image grow with each passing day. Extraction linear of the image work done by hand not only slow, but also have a lot of subjective judgement, and unfavorable follow-up study. In the study will use computer to automatically extraction linear of the image. Set up one fast, objective method; expect punishing the image in a large amount. And of far one objective linear date.
Synthetic aperture radar（SAR） image have a lot of feature include non-linear and linear, in the study will use wavelet transform and spatial correlator get the feature points in the image, and use Hough transform to connect feature points of collinear in the feature image.
The principle that wavelet transform is in regarding the part changed fast in the image as the feature part, change slower part to deny, In the study ,in fist , to produce multiscale images of one group to utilize wavelet transform. Every image of multiscale images of one group includes the high pass coefficient image of orthogonal in the basis that three, such as vertical, the level and opposite angle. And use spatial correlator to multiply different multiscale images. To get feature points in the feature image
After making the feature image, and then, Hough transform as the pure tool that extraction linear. Changes the points on the feature space into the line of parameter space, after that change the point of many line to cross on parameter space into linear on feature space, And be able to change points into linear in feature space , finish extraction linear. | en_US |