SAR Image Change Detection Based on Iterative Label-Information Composite Kernel Supervised by Anisotropic Texture

Abstract

Kernel methods with specifically designed kernel function are suitable for dealing with practical nonlinear problems. However, kernel methods have found limited applications to synthetic aperture radar (SAR) image change detection in that their performances are affected by the inherent multiplicative speckle noise of SAR images. It is known that the spatialcontextual information is helpful in suppressing the degrading effects of the noise. Therefore, a label-information composite kernel (LIC kernel) constructed on the basis of the spatialcontextual information is proposed in this paper for SAR image change detection. A typical spatial information, the output-space label-neighborhood information that is extracted using all labels in the neighborhood of each pixel, may enhance noise immunity, but with inaccurate edge locations simultaneously. Consequently, the anisotropic Gaussian kernel model is utilized for analyzing anisotropic textures of the bitemporal images, and then, a comparison scheme acting on the input-space textures of the bitemporal images is proposed to supervise the extraction of the output-space label-neighborhood information in the construction of the LIC kernel. The constructed LIC kernel is of good preservation of edge locations of changed areas as well as strong noise immunity. The LIC kernel is updated iteratively with the newest change map outputted from the support vector machine, until the change map converges. Experiments on real SAR images demonstrate the effectiveness of the LIC kernel method and illustrate that it has both strong noise immunity and good preservation of edge locations of changed areas for SAR image change detection.Index Terms-Anisotropic texture analysis, kernel method and support vector machine (SVM), label-information composite kernel (LIC kernel kernel), spatial-contextual information, synthetic aperture radar (SAR) image change detection.

SAR Image Change Detection Based on Correlation Kernel and Multistage Extreme Learning Machine

“…By deforming the basic structure of the RK function with additional information, for example, the spatial information in the output space, we propose the advanced LIC kernel function in [22]. Forsaking the former habit of deforming the RK's structure, in this paper, we construct a DCK function which is independent of the RK function.…”

“…In previous work, we propose an iterative LIC kernel [22] change detection method. The LIC kernel is constructed successfully by linearly combing the RK function with the output-space label-neighborhood information extracted under the supervision of anisotropic texture analysis.…”

“…In this part, we summarize the similarities and differences between these two methods for providing an in-depth explanation of characters of them. For more details of the LIC kernel method, please refer to [22].…”

“…The two kernel functions have strong universality for remote sensing image change detection, and the RK function has considerable adaptability to SAR images. After that, we propose a SAR image change detection method [22] based on an iterative labelinformation composite (LIC) kernel and SVM. The LIC kernel is obtained by combining the RK function with the spatial information extracted from the output labels.…”

“…The spatial-neighborhood information in the output space improves further the discriminating ability, noise immunity, and regional smoothness of MS-ELM. The proposed DCK-based MS-ELM method has a similar strategy to the LIC kernel method [22]. In this paper, an in-depth theoretical and experimental comparison between the LIC kernel and the DCKbased MS-ELM is given for showing the characters and merits of the proposed method more elaborately and completely.…”