Studies On Change Detection Methods For Remote Sensing Images

The change detection in remote sensing image is defined as the procedure of quantitatively analyzing and identifying changes occurred on the earth surface from the remote sensing images at different times. Such a problem has been played a key role in the wide range of applications in which change detection methods can be used, such as investigations of forest resources, dynamic detection of land use and land cover, assessments of environment disaster, arrangements of urban growth, and monitoring of national defense, etc. It is urgent demanded and has great potential in scientific applications.In this paper, we focus on the issues related to how to extract change information effectively from single band remote sensing images. The major works can be summarized as follows:(1) A novel method of change detection in single band remote sensing images is presented. This approach extracted the changed information by obtaining the regions of interested (ROI) in the difference image based on the method of double-threshold edge link, and modified the difference image according three different characters between the region of interested and non-interested. Finally, the ROI is classified by the threshold and the change detection map is generated. Experimental results carried out on Landsat TM/ETM+ images demonstrate that the proposed method outperforms the image difference method and the method of change detection with Markov Random Field in robust and decreasing the number of the false alarms and overall alarms.(2) A new method of change detection of remote sensing images based on Nonsubsampled Contourlet Transform (NSCT) is proposed. Firstly, the noises in detail coefficients of the two original are reduced in NSCT and the difference of detail coefficients is used as the detail coefficients of the difference image. Secondly, obtaining the approximation coefficients of the difference image based on fusion strategy to the approximation coefficients of the two original images. Finally, applying the method of inverse NSCT to the difference image approximation coefficients and detail coefficients, and the difference image is obtained, and then thresholding it based on the Kittler–Illingworth (KI) threshold selection criterion. Experiments confirm the effectiveness of the proposed technique.(3) Another novel method of change detection in SAR images based on multiscale product of wavelet transform and Principal Components Analysis (PCA) algorithm is presented. Multiscale product of wavelet transform is used, in order to avoid the affect of speckle noise in each scale of Log-ratio image. To enhance the changed information in difference image, the algorithm of PCA is used to merge every scale after noise reduction, and treat the first principle image as the newly difference image, Then produces the change detection map based on the Kittler–Illingworth (KI) threshold selection criterion to the newly difference image. Experiments carried out on two SAR images. Results are compared with two methods which based on wavelet transform and PCA, and it is proved to be superior.