Unsupervised Change Detection with Constant False Alarm Rate in Heterogeneous Remote Sensing Images Based on Copula Theory

Change detection (CD) in heterogeneous remote sensing images plays a crucial role in earth observation tasks with high timeliness. However, it is frequently unavailable for existing heterogeneous CD frameworks to control false alarm rate due to the lack of appropriate statistical modeling for heterogeneous data, leading to inaccurate identification on real changed areas overshadowed by false alarms. To address this issue, motivated by the constant false alarm rate (CFAR) strategy in traditional radar signal processing, we propose a new Copula-CFAR-CD method to theoretically control false alarm rate in heterogeneous CD. In our Copula-CFAR-CD, copula functions are trained to simulate the statistical dependence of unchanged areas between heterogeneous images. Then, CFAR thresholds are calculated based on the trained copula functions and the CD results corresponding to expectant false alarm rates are obtained. Our Copula-CFAR-CD delineates a roadmap for implementing controllable false alarms in heterogeneous CD, and experimental results on two real heterogeneous CD datasets demonstrate the effectiveness of our method.