Study And Implementation On Rapidly Preprocessing Method Of Remote Sensing Image

With the development of human science and technology, human has entered a new digital era, which remote sensing technology has also risen. Current remote sensing technology has been widely applied to all aspects of the state and society, while the mass of data it brings is unprecedented, and the reliability of the data these applications require is higher and higher. Image is an important source of human understanding the world and themselves, since it contains a lot of important information and remote sensing image processing is one of its important branches.Remote sensing image is interfered by various external factors in the imaging process, and it causes the geometric distortion, which influence the use of the quality and therefore it need geometric correction. Meanwhile, the rapid processing is also a pressing problem, as parallel technology is an effective way to solve fast image processing. This paper will first divide this process into several parts, and then use the process simulation of parallel algorithms to carry out the geometric correction while certain lines of the image block edge are redundant in order to avoid inter-process communication. At the same time we will use the separation method which has been turned out to be reliableto sample image elements.Feature is the equivalent of the result properties showed by the neighborhood of one or some of the pixels, which can keep the translation, and rotation invariance. Feature-based matching has high reliability and stability. As the SIFT feature point detection method extracts the characteristics of remote sensing images, the traditional Gaussian scope feature points form 128-dimensional SIFT feature vector, which Increase calculation of feature matching. First, this text reduces the dimension of feature vectors to 36 dimensions effectively by a ring and the Euclidean distance to match characteristics. Second, it eliminates non-matching pairs with RANSAC. Last but not least, it uses the least-square method to fit the transformation relations between the two images to be matched when the image elements are still grouped by separating line and column. At the end but not last, it tests that the method can effectively improve the efficiency of the exact match.