Remote Sensing Image Geometric Positioning Accuracy Evaluation Method Research

Nowadays more and more information can be get through the remote sensing images because of the development of remote sensing technology. The remote sensing images can be used to make topographic map, provide services for resources of motherland, agriculture, forestry and so on. However the image geometric positioning accuracy can be seriously influenced as a result of the dislocation of downloading the original data and the time code, attitude of satellite, camera installation error etc. The geometric quality of remote sensing image can be reduced. Thus the geometric positionsing accuracy evaluation of remote sensing images is important and has become one of the main research issues in remote sensing area.In the past, control points of the same locus are selected through artificial or semi-automatic methods when evaluating the geometric positioning accuracy of the remote sensing images. It was time wasting and vim wasting. Therefore, the research on how to select the control points of locus autoly, effectly and correctly is important and evaluable. The follower is the main work and achievement:(1) Algorithms including Harris, SUSAN and SIFT corner detection operators which applied to the common gray images are introdurced in this paper, also the image matching based on SURF is studied in detail in this paper. Tests on stability and efficiency of extracting the control points by the algorithms are carried on and from the analysis results, it can be seen that the SIFT operator and SURF operator are better than others, and SIFT operator is slightly greater than the SURF one. But considering that the SURF operator is much faster than the SIFT one, the SURF algorithm is choosed for further study in this paper.(2) Some mismatchings persist after matching the control points exacted by SURF algorithm by using the Euclidean distance. The geometric positioning accuracy of remote sensing images would be affected seriously. Therefore, a matching method which combines the superiorities of the speed up robust features (SURF) algorithm and the random sample consensus (RANSAC) algorithm is proposed. From the results, it can be seen that the proposed method performs well in removing the error matching. However, the matching rate is low if big differences exist in the images.(3) To solve this problem of low matching rate, the SURF algorithm is further studied in detail by adding the edge information and detail information in the neighbourhood of the feature points to the SURF descriptors. From the experimental results, it can be seen that the proposed method can be more ideal in improving the matching rate.(4) The evaluation methodology and evaluation parameters of the geometric positioning accuracy are also introduced in this paper, the geometric positioning accuracy of the remot sensing images is computed and compared with that of China Centre for Resources Satellit Data and Application. Similar results are obtained and it can be seen that the proposed metho has certain feasibility.