An Improved Change Detection Based on PCA and FCM Clustering for Earthen Ruins

Abstract

Nondestructively detecting the damage change is important for the protection of the earthen ruins. How to effectively detect the subtle changes in earthen ruins is an urgent problem to be solved. In our paper, we use the difference and the log-likelihood ratio method to create a difference image, which can effectively avoid the influence of noise. Then the orthonormal eigenvectors are extracted through principal component analysis (PCA) of non-overlapping block set to create an eigenvector space which is mapped to each vector in turn to form a feature vector space. The feature vector space is partitioned into two clusters according to the feature vector approximate degree by using fuzzy c-means (FCM) clustering. The experiment results on the images of Hanguangmen earthen ruin show that this method can find the changing area simply and efficiently.