A Siamese Network Based U-Net for Change Detection in High Resolution Remote Sensing Images

Abstract

Remote sensing image change detection (RSICD) is a technique that explores the change of surface coverage in a certain time series by studying the difference between multiple remote sensing images (RSIs) collected over the same area. Traditional RSICD algorithms exhibit poor performance on complex change detection (CD) tasks. In recent years, deep learning (DL) techniques have achieved outstanding results in the fields of RSI segmentation and target recognition. In CD research, most of the methods treat multitemporal remote sensing data as one input and directly apply DL-based image segmentation theory on it while ignoring the spatio-temporal information in these images. In this article, a new siamese neural network is designed by combing an attention mechanism (Siamese_AUNet) with UNet to solve the problems of RSICD algorithms. SiameseNet encodes the feature extraction of RSIs by two branches in the siamese network, respectively. The weights are shared between these two branches in siamese networks. Subsequently, an attention mechanism is added to the model in order to improve its detection ability for changed objects. The models are then compared with conventional neural networks using three benchmark datasets. The results show that the Siamese_AUNet newly proposed in this article exhibits better performance than other standard methods when solving problems related to weak CD and noise suppression.

A Deep Learning Model for Change Detection on Satellite Images

“…In those tables we compare the results of FC-EF-res with other architectures. The first one is Siam [3] which is a Siamese network. The second one is EF [3] which is very similar to FC-EF-res: it uses the same technique of early fusion coupled with a U-net architecture.…”

“…The first one is Siam [3] which is a Siamese network. The second one is EF [3] which is very similar to FC-EF-res: it uses the same technique of early fusion coupled with a U-net architecture. However, it is not a fully convolutional network as FC-EF-res because it has fully connected layers.…”

“…Another possibility would be to change the architecture of the network to make it invariant to permutations. A first step to achieve this is to use a Siamese network architecture, as presented by [1,3]. In a Siamese network both images pass through the same encoder separately.…”

Change Detection for High-Resolution Remote Sensing Images Based on a Multi-Scale Attention Siamese Network

“…local attention module and convolutional block attention module (CBAM), then the ASPP module is used to improve the detection effect of multi-scale change features [21].…”

“…Because existing methods fail to predict the edges and preserve the shape of the changed area from bi-temporal images, Basavaraju et al introduced a network based on an encoder-decoder architecture (UCDNet) that uses improved residual connections and a new spatial pyramid pooling (NSPP) block to obtain better prediction results while preserving the shape of changing regions [20]. Chen and Lu et al proposed Siamese-AUNet by combining a Siamese network, attention mechanism and U-Net for the detection of weakly changing objects, and the representation ability of weak features is improved by combining a non-local attention module and convolutional block attention module (CBAM), then the ASPP module is used to improve the detection effect of multi-scale change features [21].…”