PROJECT TITLE :
SAR Image Change Detection Based on Iterative Label-Information Composite Kernel Supervised by Anisotropic Texture
Kernel ways with specifically designed kernel perform are suitable for coping with sensible nonlinear issues. However, kernel methods have found restricted applications to artificial aperture radar (SAR) image modification detection in that their performances are full of the inherent multiplicative speckle noise of SAR images. It is known that the spatial-contextual information is useful in suppressing the degrading effects of the noise. Therefore, a label-info composite kernel (LIC kernel) constructed on the basis of the spatial-contextual information is proposed in this paper for SAR image modification detection. A typical spatial info, the output-space label-neighborhood data that is extracted using all labels within the neighborhood of each pixel, might enhance noise immunity, however with inaccurate edge locations simultaneously. Consequently, the anisotropic Gaussian kernel model is utilized for analyzing anisotropic textures of the bitemporal pictures, and then, a comparison scheme working on the input-space textures of the bi-temporal pictures is proposed to supervise the extraction of the output-area label-neighborhood info in the construction of the LIC kernel. The constructed LIC kernel is of good preservation of edge locations of changed areas with robust noise immunity. The LIC kernel is updated iteratively with the latest modification map outputted from the support vector machine, till the amendment map converges. Experiments on real SAR images demonstrate the effectiveness of the LIC kernel method and illustrate that it's both robust noise immunity and smart preservation of edge locations of modified areas for SAR image change detection.
Did you like this research project?
To get this research project Guidelines, Training and Code... Click Here