Research On Change Detection Of Remote Sensing Images Based On Treelet Transform

The change detection of remote sensing images is a process of identifying statedifference of multiple remote sensing images acquired from the same area at differenttimes. With the fast development of remote sensing technology and informationprocessing technology, the change detection of remote sensing images has become animportant direction in the study of remote sensing images, and has been widely used invarious fields, such as land use and land cover, urban arrangement, environmentmonitoring, crop monitoring, military reconnaissance and so on.Taking treelet transform as the core, we carry out the research on how to extractchange information from single-band two-temporal remote sensing images, and hascompleted three works as the follows:(1) A change detection method of remote sensing images based on treelet denoisingin curvelet domain is proposed. Firstly, using treelet adaptively denoise the logarithmicdifference image in curvelet domain. Secondly, computing the classification thresholdby the low-frequency information in curvelet domain. And then thresholding the directdifference image and reconstructed image. Finally, making a fusion to obtain the changedetection results. The effectiveness of the proposed method is verified by theexperiment on a set of simulation data and four sets of real remote sensing data.(2) A change detection method of remote sensing images based on treelet andimage fusion is proposed. Firstly, obtaining a precise detection result by statistics meanvalue classification and adaptive spatial information filling. Secondly, obtaining a roughlocation result by treelet fusion, k-means, and mathematical morphology post-treatment.Finally, obtaining the change detection result by making fusion. The effectiveness of theproposed method is verified by the experiment on a set of simulation data and four setsof real remote sensing data.(3) A change detection method of remote sensing images based on treelet anddirection adaptive filtering is proposed. Firstly, using direction adaptive filtering todenoise difference image, and then making a fusion by treelet. Secondly, thresholdingthe fusion image by adaptive threshold classification to get a classification image.Finally, making a post-process based on area threshold to generate change detectionresult. The effectiveness of the proposed method is verified by the experiment on a setof simulation data and four sets of real remote sensing data.