SAR image can be matched study
|School||Huazhong University of Science and Technology|
|Course||Communication and Information System|
|Keywords||Image Matching SNR Correlation coefficient Accuracy estimates|
Image matching is a very critical navigation links instant access to real-time by drawing out machine with pre-stored in the reference graph matching, anti-carrier aircraft current position is calculated, calculating and planning path deviation for track correction. Image registration directly affect the accuracy of the navigation accuracy. And visible light imaging in different ways, SAR is slant range imaging, and the ground elevation and downs of a great relationship, in the case of flat ground, ground objects in real-time on the chart with reference to FIG conformation same ups and downs occur when the local surface, will cause the image differences occur, when the difference reaches a certain extent it will cause difficulty for the registration, so the analysis of the SAR image can be matched, for accurate navigation has important significance. This article about the SAR image matching research can be divided into the following two aspects: Analysis of accuracy of registration and registration accuracy of quantitative estimates. This paper is based on the relevant area normalized cross-correlation algorithm, on this basis, matching the image can be studied. This paper first introduces the domestic and foreign research status on image matching; Second, in the basic principles of SAR imaging, SAR image distortion types and matching algorithm After a brief introduction, factors affecting image registration depth discussions, analysis of indicators and statistical characteristics of the image surface features related indicators of significance for the description of image features, and the indicators for impact on accuracy reasons, respectively, in a single experiment verified under experimental and statistical characteristic indexes relationship with registration accuracy, and the experimental results are given a reasonable explanation; Then, the amount of information and analysis of the image distortion on the combined effects on accuracy, obtained distortion on registration accuracy was greater than the impact of the amount of information, as well as a small amount of information distortion is the best conclusion; Finally, the bridge by the correlation coefficient, the image registration accuracy of the SNR and the link, the SNR and the first correlation coefficient derived quantitative expression, on the basis of analysis of the correlation coefficients between the accuracy and register , proposed a feature based on the correlation coefficient distribution registration accuracy quantitative estimation method. The method correlation coefficient distribution peak width as a threshold value for the most relevant surface around the point of a neighborhood's correlation coefficient to determine, identify areas may result in mismatches, suspect areas most value from that point to the most distant To estimate the distribution standard deviation for the estimation method, the paper carried a large number of experiments with noise data to prove the accuracy of the method. Finally, a summary of the full text, pointing out the need to consider aspects as well as future research work.