Change Detection For Remotely Sensed Imagery

With the rapid development of the remote sensing technology, remotely sensed images processing technologies have been widely applied in military, politics and our daily life. The remote sensing change detection technology can be used to analyze the remote sensing images captured in the same area but at different times to acquire the changes. The technology can be used in many important fields, such as the urban planning, land monitoring, disaster evaluating, etc. Especially, synthetic aperture radar (SAR) images can be captured without regarding weather condition and illumination, so that it will be more promising in applications. In this thesis, we conclude some related remote sensing change detection technologies and propose two change detection algorithms:In order to deal with the optical images, we propose the remote sensing change detection algorithm based on Lo smoothing and superpixel techniques. In this algorithm, Lo smoothing can simultaneously sharpen major edges and reduce noise effect. Superpixel technique is applied to obtain roughly uniform superpixels while preserving local image boundaries. Then, we calculate the change degree for each superpixel to classify the superpixels into changed and unchanged class.In order to deal with the SAR images change detection problem, we propose the change detection algorithm based on patch similarity and Lo soft segmentation. We propose the difference image for SAR image change detection based on patch similarity to face up with the lack of the suitable difference image. Meanwhile, we propose the difference image analysis method based on Lo soft segmentation to separate images into change and unchanged regions. Compared with the traditional Chan-Vese segmentation model, our algorithm is with high efficiency.In conclusion, we propose two change detection algorithms in order to deal with optical and SAR images respectively. Experimental results show that our algorithms have a certain effect on change detection problems.