Several Key Technological Research Of Change Detection Of Remote Sensing Images Base On Unmanned Aircraft

Remote sensing image from unmanned aerial vehicle(UAV), because of its high spatialresolution and high temporal resolution characteristics, is increasingly becoming an integral partof the data source feature change detection. Currently, change detection techniques of the UAVremote sensing image are basically at the artificial visual interpretation stage, lacking ofautomation. This article focuses on the key technologies to conduct research around changedetection techniques of the UAV remote sensing image, the main work is as follows:(1) Mutual information constraints of background was introduced, then improved SIFTregistration algorithm. For the reliability of SIFT feature matching, In outside of local feature ofSIFT descriptor outside, establish a semi-global background ring, and the mutual information ofbackground ring is embedded into similarity measure constraint feature matching. This modifiedalgorithm effectively improved the correct rate of matching, and there was a smaller amount ofcalculation compared with the global alignment algorithm;(2) Multi-scale remote sensing image change detection method based on local mixedinformation was proposed. from the grayscale image change detection method of Pseudo-changebased on mixed information combined with wavelet transform decomposition of multi-scaleimage, multi-scale mixed information change detector was constructed. The experimental resultsshow that, compared to the mixed information change detection method, an improved methodretains more details of the changes to reduce the area of the pseudo-change to improve therobustness of the change detection;(3) Fuzzy clustering change detection method of Multi-scale image of abnormalaccumulation was proposed. In order to overcome geometric registration error of color image, asymmetric CC cumulative abnormal change detection was constructed, using FCM clusteringmethod to classify the the accumulated anomaly detection Figure on the basis of symmetry theChrono-chrome (CC) detection method,. The experimental results show that, compared with thechange detection method of the symmetrical CC, the improved method can better overcome theimage registration error, and the ability to detect the true color changing region wasunsupervised.