An Efficient Method for Typical Target Feature Extraction and Recognition in SAR Images
Lin Hong-jin · Journal of Sichuan University · 2005
In synthetic aperture radar(SAR) imaging,man-made and natural object recognition and classification is necessary.Object's feature extraction is a key component in automatic target recognition.In most cases,color,texture and shape are adopted as image's features which are not functioned well in SAR images due to a lot of noise in the background and blurry boundary. Authors propose a method for object feature extraction and recognition of SAR images.In the method,based on the theory of SAR imaging,the image intensities are segmented and used as the masks of searching regions.Then,the image is searched sparsely and gray feature,ellipse vector feature,and grid feature are extracted.Finally,the typical objects in the image are classified and recognized.Experiments show that the method is preferable efficient and satisfactory.