Component Analysis-Based Change Detection for Ocean Floor Multitemporal Imagery of Synthetic Aperture Sonar

In this article, we present a novel approach for ocean floor change detection in multitemporal synthetic aperture sonar (SAS) imagery. The process involves detecting changes from pairs of multitemporal SAS images, extracting anomaly features, and reducing false alarms, all of which are complicated by random speckle patterns, environmental variability, and platform instabilities. These factors make target detection and classification challenging. This article introduces a comprehensive automated change detection (ACD) framework that incorporates coregistration of reference and repeat-pass data sets, incoherent change-map generation, anomaly detection, and advanced false alarm reduction techniques. Two false alarm reduction methods are proposed: one based on a reduction step using zero-detect thresholding, and the other employing prioritized statistical techniques through principal component analysis (PCA) and independent component analysis (ICA). The zero-detect threshold method dynamically adjusts thresholds based on environmental statistics, while PCA reduces false alarms by analyzing eigenvalues to identify significant components. ICA incorporates higher order statistics, such as skewness and kurtosis, to prioritize independent components for improved target detection. Experimental results using real SAS data sets demonstrate the efficiency of the proposed methods, achieving significant noise reduction and enhanced target detection. The integration of PCA and ICA improves target detection accuracy by mitigating the effects of noise, environmental disturbances, and repetitive patterns on the ocean floor. This research advances ocean floor monitoring by providing a robust, adaptable ACD process. Future work will explore the extension of these techniques to multisensor systems and investigate coherence-based methods for enhanced target classification, addressing the growing needs of underwater surveillance and oceanographic research.