MAEANet: Multiscale Attention and Edge-Aware Siamese Network for Building Change Detection in High-Resolution Remote Sensing Images

Abstract

In recent years, using deep learning for large area building change detection has proven to be very efficient. However, the current methods for pixel-wise building change detection still have some limitations, such as a lack of robustness to false-positive changes and confusion about the boundary of dense buildings. To address these problems, a novel deep learning method called multiscale attention and edge-aware Siamese network (MAEANet) is proposed. The principal idea is to integrate both multiscale discriminative and edge structure information to improve the quality of prediction results. To effectively extract multiscale discriminative features, we design a contour channel attention module (CCAM) that highlights the edge of the changed region and combine it with the classical convolutional block attention module (CBAM) to construct multiscale attention (MA) module, which mainly contains channel, spatial and contour attention mechanisms. Meanwhile, to consider the structure information of buildings, we introduce the edge-aware (EA) module, which combines discriminative features with edge structure features to alleviate edge confusion in dense buildings. We conducted the experiments using LEVIR-CD and BCDD datasets. The proposed MA and EA modules can improve the F1-Score of the basic architecture by 1.13% on the LEVIR CD and by 1.39% on the BCDD with an accepted computation overhead. The experimental results demonstrate that the proposed MAEANet is effective and outperforms other state-of-the-art methods concerning metrics and visualization.

Automatic Point Cloud Colorization of Ground-Based LiDAR Data Using Video Imagery without Position and Orientation System

“…Due to the phenomenon of blurred edges in video images, normalized Zernike moments are used to describe the feature region. Low-order Zernike moments describe the overall shape characteristics of the image, while high-order Zernike moments reflect the texture details of the image [33][34][35]. Normalized Zernike moments are invariant to rotation, translation, and scaling and can serve as the determining factor for registration.…”

“…Therefore, in this paper we propose a registration method based on normalized Zernike moments, which is also a point feature-based registration method. Low-order Zernike moments are mainly used to describe the overall shape characteristics of the image [33], while high-order Zernike moments are mainly used to reflect texture details and other information of the image [34]. Normalized Zernike moments can reflect features in multiple dimensions [35], achieving good results even for low-quality images captured by a GoPro action camera and point cloud intensity images.…”

Progressive Difference Amplification Network With Edge Sensitivity for Remote Sensing Image Change Detection

“…In terms of edge detail modeling, Chen et al [41] proposed an edge-aware module (EAM) based on dilated convolution and gated attention, and embedded EAM in each encoder block to refine the edge details. Yang et al [42] designed the contour channel attention module (CCAM) to refine the high-resolution decoded features and highlight the edges of changed areas, where CCAM uses superpixel object segmentation as a reference to enhance the internal consistency of the changed area. Xia et al [14] constructed an edge detection branch to obtain multiscale edge features for masking decoded difference features.…”

Multi-Scale Discrete Cosine Transform Network for Building Change Detection in Very-High-Resolution Remote Sensing Images

“…The method can effectively enhance the building change extraction ability by introducing the edge of the building to guide the recurrent convolutional neural network. Yang et al devised a multi-scale attention and edge-aware Siamese network in [24] for BCD. These methods have proved that the introduction of edge information can improve the performance of BCD.…”