| As the developing of space technology, the resolution of the remote sensing image increases rapidly. It becomes feasible that the high resolution remote sensing image can be used to survey map of large scale from 1:25000 to 1:10000 and update geography information, which used airphoto in the past.It is inevitable that the satellite remote sensing images have geometry distortion caused by some kinds of exterior and interior factors. Some methods of rectification have to be taken to correct the distortion, i.e. fulfilling the rectification of an image by using a group of control points on the ground. The rectification was preformed by using some kinds of professional software, such as ERDAS software system. In order to get the coordinate of feature points, the operator moves his mouse on the computer screen and locates the control points one by one.Currently used rectification method are not satisfying. They have some disadvantages, such as heavy workload, inefficiency, difficulty in finding out a great deal of feature points, and so on. In the case of lacking of ground control points, the rectification is far more difficult. At the same time, there are large numbers of vector maps with various scales. As a kind of fundamental geography information, it contains so many feature points with correct geographic coordinate, and does not be used in the rectification.The form and content of remote sensing image and vector map are quite different. Vector map represents shape, size, and other attributes of those objects in the image. How to make good use of vector maps in the rectification is what discussed in this paper.High resolution remote sensing image can provide abundant information of the objects in spectrum, shape, structure, texture, and so on. Vector map includes plenty of feature points and feature lines. By means of computer automatic processing, the feature points and lines in the maps and the equivalents in the images can be found out efficiently and automatically. Using the "edge feature matching" method, which was brought forward in this paper, we can construct "vector feature template" and "image feature window" to establish the corresponding relationship of them. Finally, according to the rectification polynomial, the remote sensing image was resampled and rectified.High resolution remote sensing image can make high precision matching. At the same time, its abundant details will increase the probability of error matching. When matching with vector map, the background information in the remote sensing image can not be used. In this case, the probability of error matching is higher than matching between two raster images. In this method, we can select 4 to 6 points to get the rough rectification relationship. In this way, we can accelerate the processing, reduce the searching area of matching, decrease the amount of calculation, and enhance the stability of this algorithm. Afterprocessing recursively, we can get the optimal rectify polynomial.So, In this paper, we discuss a new method of fast geometric rectification on high resolution remote sensing image. With the comprehensive utilization of several kinds of technology, such as digital image processing and analyzing, image enhancement , image segmentation, edge feature matching and image registration, remote sensing image can be rectified by means of performing the registration between image and vector map. Then, the image has new correct geographic coordinate as same as the map. The resourceful vector map data can be made use of as its objects matched on the remote sensing image. This method has resolved the problem of the matching of feature points between vector map and raster image. It has found out a new way of getting pairs of feature points automatically. The aim of the new method is to increase the degree of automatization and precision of high resolution image rectification, transact the image more efficiently, and improve the effect of rectification. Under the environment of VISUAL C++ .NET , the programme of quickly rectification was developed. The result of experiment shows that the idea presented in the paper is reasonable and applicable.It is necessary for further researching on matching algorithms and image rectification in sub-regions in the future. |
