Registration And Change Detection For Remote Sensing Images By Multiscale Geometric Analysis And Optimization Of Natural Evolution

Registration and change detection for Remote Sensing images are important parts of processing of Remote Sensing images.Accuracy of registration is prior to the precision of change detection.Firstly,a despeckling method for Synthetic Aperture Radar(SAR) images was proposed. Secondly,registration method by gray level of images and the other method by features of images were both studied.Thirdly,in order to reduce the impact with noises in change detections novel change detection method was proposed. A approach distribution free was proposed for change detection of SAR images.The speckle degrades the quality of the images and makes interpretations,analysis and classifications of SAR images harder.Therefore,speckle reduction is necessary prior to the processing of SAR images.Wedgelet approximation can obtain edge information in images effectively while smoothing the texture regions.A method for SAR images using Wedgelet combined with Dual tree complex wavelet transform(DTCWT) was proposed.Furthermore,an edge-enhanced local mean and median filter by MSP-ROA edge detector is added to smooth SAR speckle noise while preserving edges.The experimental results performed on SAR images have shown that it does perform better in both noise smoothing and edge preservation.Mutual information with multi-agent optimization algorithm was proposed for SAR image registration.The normalized mutual information measure based on gray images was used as a matching criterion without pretreatment of images.The final result of the registration was obtained by intelligent optimization algorithm.The results of experi-ments show that the method is good for registration of SAR images.In order to improve the accuracy of registration for multi-modal remote sensing images,a algorithm by best key points matching was proposed.Firstly control points of test image and the reference image were setted.The best matching of key points was determined using Difference of Gaussian(DOG).Secondly,a rough registration of the test image and reference image was getted by using projection transform and linear least square algorithm.Finally the algorithm automatically adjusts the control points by sub-pixel step according to the registration error and achieves the sub-pixel registration.In order to reduce the impact with noises in change detection for remote sensing images,a novel change detection method was proposed,which based on immune clone algorithm and wavelet transform. Firstly,the multi-scale and low-pass smoothing characteristics of wavelet transform were utilized to construct multi-layered difference images. Secondly,the time domain deviations caused by operation of zero insertion and image convolution in wavelet transform were corrected by immune clonal algorithm that the initial segmentation results obtained using Rayleigh-Gauss model were matched by secondary linear interpolation operation.Finally,the change detection was accomplish-ed by image fusion. Simulation results show that the algorithm can not only reduce the image noises,but also suppress the image deviations caused by wavelet transform effectively.The accuracy of change detection is improved significantly.Most existing change detection algorithms obtain the change area by image segmentation of the difference image.However,the parameters of distribution model must be setted.If parameters does not match,precision of the change detection will reduce.Many improved algorithm were proposed,but initial segmentation results were still limitted by the parameters.Addressing above,a method distribution free was studied. The change result was obtained by a clustering method with minimizing mean square error(MSE) using Biogeography-based Optimization(BBO) algorithm.The results of experiments show that the method is good for change detection of Remote Sensing images.