An Integrated Change Detection Method Based on Spectral Unmixing and the CNN for Hyperspectral Imagery

Hyperspectral remote sensing image (HSI) include rich spectral information that can be very beneficial for change detection (CD) technology. Due to the existence of many mixed pixels, pixel-wise approaches lead considerable errors in resulting CD map. The unmixing (SU) method is a potential solution this problem, as it decomposes pixels into set fractions land cover. Subsequently, map created by comparing abundance images. However, based only on images through SU method, they are unable effectively provide detailed information. Meanwhile, features cannot sufficiently extracted traditional sub-pixel framework, which leads poor result. To address these problems, paper presents an integrated multi-endmember unmixing, joint matrix and CNN (MSUJMC) HSI. Three main steps considered accomplish task. First, considering endmember variability, more reliable obtained (MSU). Second, original incorporated with using (JM) algorithm temporal spatial cover characteristics. Third, efficiently extract better handle fused multi-source information, convolutional neural network (CNN) introduced realize high-accuracy proposed has been verified simulated real multitemporal HSI datasets, multiple changes. Experimental results verify effectiveness approach.