Unsupervised Change Detection in Hyperspectral Images using Principal Components Space Data Clustering

Abstract

Change detection of hyperspectral images is a very important subject in the field of remote sensing application. Due to the large number of bands and the high correlation between adjacent bands in the hyperspectral image cube, information redundancy is a big problem, which increases the computational complexity and brings negative factor to detection performance. To address this problem, the principal component analysis (PCA) has been widely used for dimension reduction. It has the capability of projecting the original multi-dimensional hyperspectral data into new eigenvector space which allows it to extract light but representative information. The difference image of the PCA components is obtained by subtracting the two dimensionality-reduced images, on which the change detection is considered as a binary classification problem. The first several principal components of each pixel are taken as a feature vector for data classification using k-means clustering with k=2, where the two classes are changed pixels and unchanged pixels, respectively. The centroids of two clusters are determined by iteratively finding the minimum Euclidean distance between pixel’s eigenvectors. Experiments on two publicly available datasets have been carried out and evaluated by overall accuracy. The results have validated the efficacy and efficiency of the proposed approach.

Temporal Trends and Spatial Relationships of Radioactive Isotopes (I-131, Cs-134, and Cs-137) in Response to Nuclear Events: A Comprehensive Analysis Using Time Series Graphs, Regression, and Multivariate Techniques

“…These eigenvectors are the principal components that capture the maximum variance in the data. The original data is projected onto the selected k principal components to obtain the reduced-dimensional representation [59]. For the i-th data point, the reduced representation i z is calculated as Eq.…”

Qualitative and quantitative analysis of artificial neural network-based post-classification comparison to detect the earth surface variations using hyperspectral and multispectral datasets

“…In the previous literature, various change detection techniques were developed and improved to enhance the utilization of hyperspectral imagery, such as convolutional sparse analysis, temporal spectral unmixing, 41 recurrent 3D fully convolutional networks, 42 convolutional long short-term memory, 42 modified U-Net model, 43 principal component analysis, and spectral correlation angle. 44 But the processing of hyperspectral bands with such advanced change detection models may be difficult due to the presence of a large number of spectral bands, which may limit its applicability in various applications. Moreover, the existence of radiometric/atmospheric errors in the hyperspectral dataset may make it difficult for change detection models to identify subtle changes.…”

“… – 40 However, it has been observed that change detection via hyperspectral imaging delivers refined information in various narrow spectral bands that may not be possible with change detection via multispectral imaging. In the previous literature, various change detection techniques were developed and improved to enhance the utilization of hyperspectral imagery, such as convolutional sparse analysis, temporal spectral unmixing, 41 recurrent 3D fully convolutional networks, 42 convolutional long short-term memory, 42 modified U-Net model, 43 principal component analysis, and spectral correlation angle 44 . But the processing of hyperspectral bands with such advanced change detection models may be difficult due to the presence of a large number of spectral bands, which may limit its applicability in various applications.…”