Semantic change detection (SCD) has gradually emerged as a prominent research focus in remote sensing image processing due to its critical role in earth observation applications. In view of its powerful semantic-driven feature extraction capability, the Segment Anything Model (SAM) has demonstrated its suitability across various visual scenes. However, it suffers from significant performance degradation when confronted with remote sensing images, especially those containing various ground objects that possess significant inter-class similarity and substantial intra-class variations. To address the above issues, we propose SCD-SAM, aiming to leverage the potent visual recognition capabilities of SAM for enhanced accuracy and robustness in SCD. Specifically, we introduce a contextual semantic change-aware dual encoder that combines MobileSAM and CNN to extract progressive semantic change features in parallel, and inject local features into the MobileSAM encoder through depth feature interaction to compensate for the Transformer’s limitations in perceiving local semantic details. Besides, in order to utilize the strong visual feature extraction capability of MobileSAM in remote sensing images, we propose a semantic adaptor that aggregates semantic-oriented information about changing objects. To better integrate the extracted contextual semantic information, we devise a progressive feature aggregation dual decoder that aggregates binary change features and semantic change features respectively, alleviating the semantic gap across different scales. The quantitative and visual results show that SCD-SAM outperforms the state-of-the-art SCD methods on publicly open SCD datasets (e.g., SECOND-CD and Landsat-CD). The code will be made available at https://github.com/yzygit1230/SCD-SAM.

Breast cancer is one of the most prevalent diseases for women worldwide. Early and accurate ultrasound image segmentation plays a crucial role in reducing mortality. Although deep learning methods have demonstrated remarkable segmentation potential, they still struggle with challenges in ultrasound images, including blurred boundaries and speckle noise. To generate accurate ultrasound image segmentation, this paper proposes the Edge-Aware Multi-Scale Group-Mix Attention Network (EMGANet), which generates accurate segmentation by integrating deep and edge features. The Multi-Scale Group Mix Attention block effectively aggregates both sparse global and local features, ensuring the extraction of valuable information. The subsequent Edge Feature Enhancement block then focuses on cancer boundaries, enhancing the segmentation accuracy. Therefore, EMGANet effectively tackles unclear boundaries and noise in ultrasound images. We conduct experiments on two public datasets (Dataset- B, BUSI) and one private dataset which contains 927 samples from Renmin Hospital of Wuhan University (BUSIWHU). EMGANet demonstrates superior segmentation performance, achieving an overall accuracy (OA) of 98.56%, a mean IoU (mIoU) of 90.32%, and an ASSD of 6.1 pixels on the BUSI-WHU dataset. Additionally, EMGANet performs well on two public datasets, with a mIoU of 88.2% and an ASSD of 9.2 pixels on Dataset-B, and a mIoU of 81.37% and an ASSD of 18.27 pixels on the BUSI dataset. EMGANet achieves a state-of-the-art segmentation performance of about 2% in mIoU across three datasets. In summary, the proposed EMGANet significantly improves breast cancer segmentation through Edge-Aware and Group-Mix Attention mechanisms, showing great potential for clinical applications.

Building change detection (BCD) is essential for urban dynamic measurement. Deep learning has demonstrated significant potential in image processing, providing powerful feature extraction capabilities for BCD tasks. However, existing methods do not adequately mine multiscale feature information and ignore the importance of multiscale feature alignment, leading to an inadequate representation of the internal structure. Therefore, we propose a hybrid attention-aware transformer network (HATNet) designed to effectively extract and interact with multiscale context information. Specifically, HATNet first incorporates a hybrid attention-aware feature extractor (HAFE) module that integrates self-attention (SA) and coordinate-attention (CA) to effectively extract complementary multiscale features. The SA establishes long-range dependencies between multiscale features, while the CA captures spatial dependencies and preserves positional details. Then, we devise a building saliency detection enhancement (BSDE) module that utilizes three independent channels to facilitate the identification and localization of changed buildings, fostering a symbiotic relationship between unchanged and changed areas. Furthermore, we adopt a coarse-to-fine feature interaction (CFFI) module to progressively fuse multiscale features using a hierarchical strategy. In order to better locate the changed detail features, we introduce a global feature alignment (GFA) module to achieve global multiscale feature alignment. HATNet surpasses eight BCD methods on LEVIR-CD, WHU-CD, and S2Looking-CD datasets. The experimental results provide compelling evidence of our method to precisely detect and measure building changes. The codes are available at https://github.com/yzygit1230/HATNet .

Cropland resources are essential for the provision of food production, which is one of the most fundamental needs of human life. Change detection (CD) technology enables the dynamic monitoring of high-resolution cropland resource images acquired through remote sensing satellite sensors. However, current CD methods are not capable of extracting meaningful change information from dense and continuously distributed cropland. Additionally, the common feature fusion processing often results in information redundancy and the loss of key features. Therefore, we propose a Cross-Attention Guided Group Aggregation Network (CAGNet) to achieve effective cropland CD. Specifically, we adopt a cross-attention (CA) module to enhance the capability of extracting and characterizing the features of the changed region, reducing the influence of noise and pseudo-change on CD. To alleviate the loss of key information during the multi-scale feature fusion process, and thus improve CD performance, we design a group aggregation (GA) module that gradually groups and aggregates the bitemporal features from coarse to fine. Finally, we use a fully convolutional network to obtain detailed CD results. Furthermore, we demonstrate the effectiveness of knowledge transfer in the field of CD. It allows the models to obtain the underlying mechanisms and characterization capabilities of changed features on the building CD dataset in advance, which significantly improves the performance of various methods on the cropland CD dataset. The experimental results show that CAGNet’s quantitative metrics results on the cropland dataset (CL-CD) outperform the other ten benchmarked methods, achieving an F1-score of 79.53%.

Building change detection (BCD) using high-resolution remote sensing images aims to identify change areas during different time periods, which is a significant research focus in urbanization. Deep learning methods are capable of yielding impressive BCD results by correctly extracting change features. However, due to the heterogeneous appearance and large individual differences of buildings, mainstream methods cannot further extract and reconstruct hierarchical and rich feature information. To overcome this problem, we propose a progressive context-aware aggregation network combining multi-scale and multi-level dense reconstruction to identify detailed texture-rich building change information. We design the progressive context-aware aggregation module with a Siamese structure to capture both local and global features. Specifically, we first use deep convolution to obtain superficial local change information of buildings, and then utilize self-attention to further extract global features with high-level semantics based on the local features progressively, which ensures capability of the context awareness of our feature representations. Furthermore, our multi-scale and multi-level dense reconstruction module groups extracted feature information according to pre- and post-temporal sequences. By using multi-level dense reconstruction, the following groups are able to directly learn feature information from the previous groups, enhancing the network’s robustness to pseudo changes. The proposed method outperforms eight state-of-the-art methods on four common BCD datasets, including LEVIR-CD, SYSU-CD, WHU-CD, and S2Looking-CD, both in terms of visual comparison and objective evaluation metrics.

Road damage detection is essential to the maintenance and management of roads. The morphological road damage contains a large number of multi-scale features, which means that existing road damage detection algorithms are unable to effectively distinguish and fuse multiple features. In this paper, we propose a dense multiscale feature learning Transformer embedding cross-shaped attention for road damage detection (DMTC) network, which can segment the damage information in road images and improve the effectiveness of road damage detection. Our DMTC makes three contributions. Firstly, we adopt a cross-shaped attention mechanism to expand the perceptual field of feature extraction, and its global attention effectively improves the feature description of the network. Secondly, we use the dense multi-scale feature learning module to integrate local information at different scales, so that we are able to overcome the difficulty of detecting multiscale targets. Finally, we utilize a multi-layer convolutional segmentation head to generalize the previous feature learning and get a final detection result. Experimental results show that our DMTC network could segment pavement pothole patterns more accurately and effectively than other methods, achieving an F1 score of 79.39% as well as an OA score of 99.83% on the cracks-and-potholes-in-road-images-dataset (CPRID).