Abstract
AbstractBuilding change detection in high-resolution remote sensing images is very important for illegal building management and urban supervision. Recently, with the development of neural network and the increase of RS data, there are more and more change detection methods based on deep learning. Most of the existing change detection algorithms based on deep differential feature analysis which detect all semantic changes in two-temporal images, not specifically designed for building change detection and unable to give an accurate mask for building changes area. In this paper, we propose a Siamese U-net with attention mechanism for building change detection in high-resolution bi-temporal remote sensing images. By introducing scene-level building segmentation, we improve the boundary integrity and internal compactness of the final changed building. Our method was applied to WHU dataset and have outstanding building change detection results.KeywordsChange detectionDeep learningFully convolutional Siamese networkRemote sensing image processing
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have