Hybrid Linear Discriminant Analysis (HLDA) based Dimensionality Reduction and Unary Potential based on Fuzzy Possibilistic C-Means (FPCM) Clustering for Remote Sensing Images

Abstract

High spatial resolution (HSR) remote sensing images provide detailed geometric information about land cover. As a result, it is possible to detect more subtle changes with the help of HSR images. However, due to the increased spatial resolution and the limited spectral information, it is difficult to identify the real changes only through the spectral feature of the image. Recent work Hybrid Conditional Random Field (HCRF) model is proposed which fully explore the spectral–spatial information and improve the change detection performance for HSR images. But here the effect of the unary potential, which plays an important role in the HCRF model, will be further investigated in this work. In this paper, Unary Potential is performed based on Fuzzy Possibilistic C-Means (FPCM) clustering algorithm and it is used for an object-based technique. In order to reduce the dimension of the spectral–spatial information, Hybrid Linear Discriminant Analysis (HLDA) is proposed in this work. In the proposed work introduce a new LDA matrix computation that incorporates confusability information between classes into the transform. It introduces a weighting factor for each pairwise scatter that enables to integrate the confusability information into the between class covariance matrix. In the proposed method, the spectral discriminative information of a single pixel is extracted by the unary potential, which is modeled using a FPCM clustering method to make an initial separation of changed and unchanged pixels. The pairwise potential then considers the contextual information of adjacent pixels to favor spatial smoothing. Using the accurate unary potential and reasonable pairwise potential for HSR images, the CD result can be greatly improved to avoid the over smoothing problem.