Semi-Supervised Change Detection Method Using Phase Congruency and Local Binary Pattern (PC-LBP)

Abstract

Several feature extraction methods are available for the application of change detection using remote sensing images. In this paper an efficient feature extraction technique is proposed by combining the phase congruency (PC) and local binary pattern (LBP) to extract the features in the images. The congruency amplitude and orientation is first determined for the two-multi temporal images and the LBP features are extracted. From the feature vectors of two images, neural networks (NN) classify the region into changed and unchanged. The deterministic method developed in this paper exploits improved accuracy in the change detection results. From the performance evaluation, it is noted that proposed method shows high accuracy and better kappa coefficient, which yields good change detection results.