Research And Application Of Images Change Detection Algorithm Based On NSCT And NSST

With the development of remote sensing technology, change detection technology of remote sensing image has become a hotspot in the research of the remote sensing images. It is widely used in many fields such as monitoring the changes of resources in forest, land and wate, assessment of natural disasters on fires, earthquakes and floods, agricultural resources exploration, military deployment and the dynamic monitoring of military target and so on.This paper mainly studies on the change detection algorithm of multi-temporal remote sensing images and its application in medical images and remote sensing images where alien species invade.(1) An unsupervised Image change detection algorithm based on fractal dimension and FCM clustering is proposed. The algorithm combines Nonsubsampled Contourlet Transform(NSCT) with fractal dimension, overcoming the wavelet transform’s disadvantages of down-sampling and can only decompose in the horizontal and diagonal directions. The difference map is generated by fractal dimension method, which has the characteristics of combining the spatial information and gray level information of the image. While in the calculation of the fractal dimension of the image, the speckle noise can be processed directly. The proposed algorithm can well reduce the noises.The fuzzy c-means(FCM) clustering algorithm can cluster the difference image into two clusters and get the good result.(2) A new change detection approach based on neighborhood ratio difference image and FCM clustering is proposed. The difference map is generated based on the principle of neighborhood ratio, which can well solve the problem that useful change information is difficult to extract. The problem is often caused by radiation difference, registration error and the improper selection of the difference image’s classification threshold. The difference map is transformed into NSCT domain and conducted the corresponding denoising processing. In this way the signal to noise ratio is improved under a large number of change information. The new approach is applied to the remote sensing satellite image which is invaded by invasive plants. The experiment proves that the proposed approach is effective in dynamic monitoring of plant invasion.(3) A novel change detection approach based on the neighborhood ratio difference image and the Nondownsampling Shearlet Transform(NSST) is proposed. It is applied to detect changes in white matter lesions of MR medical images. Experimental results show that the approach can accurately detect changes in white matter lesions part and realize automatic, rapid detection of lesion areas. The detection results have important significance in clinical judgment of white matter lesions trends and judgment of the clinical auxiliary diagnosis and prognosis.