Unsupervised Change Detection for VHR Remote Sensing Images Based on Temporal-Spatial-Structural Graphs

With the aim of automatically extracting fine change information from ground objects, detection (CD) for very high resolution (VHR) remote sensing images is extremely essential in various applications. However, increase spatial resolution, more complicated interactive relationships evident diversity spectra, and severe speckle noise make accurately identifying relevant changes challenging. To address these issues, an unsupervised temporal-spatial-structural graph proposed CD tasks. Treating each superpixel as a node graph, structural objects presented by parent–offspring with coarse segmented scales introduced to define temporal-structural neighborhood, which then incorporated neighborhood form neighborhood. The graphs defined on such neighborhoods extend range among nodes two dimensions three dimensions, can perfectly exploit contextual bi-temporal images. Subsequently, metric function designed according spectral similarity between measure level changes, reasonable due comprehensive utilization information. experimental results both VHR optical SAR demonstrate superiority effectiveness method.