Semantic segmentation from very fine resolution (VFR) urban scene images plays a significant role in several application scenarios including autonomous driving, land cover classification, urban planning, etc. However, the tremendous details contained in the VFR image, especially the considerable variations in scale and appearance of objects, severely limit the potential of the existing deep learning approaches. Addressing such issues represents a promising research field in the remote sensing community, which paves the way for scene-level landscape pattern analysis and decision making. In this paper, we propose a Bilateral Awareness Network which contains a dependency path and a texture path to fully capture the long-range relationships and fine-grained details in VFR images. Specifically, the dependency path is conducted based on the ResT, a novel Transformer backbone with memory-efficient multi-head self-attention, while the texture path is built on the stacked convolution operation. In addition, using the linear attention mechanism, a feature aggregation module is designed to effectively fuse the dependency features and texture features. Extensive experiments conducted on the three large-scale urban scene image segmentation datasets, i.e., ISPRS Vaihingen dataset, ISPRS Potsdam dataset, and UAVid dataset, demonstrate the effectiveness of our BANet. Specifically, a 64.6% mIoU is achieved on the UAVid dataset.

Extreme weather has been more frequent in recent years. Urban agglomerations, as areas with a high density of human activities, have been plagued by storm flooding. Historically, the main focus of attention on flood control in urban agglomerations has gradually shifted from underground pipe networks to the impervious surface, reflecting profound changes in the influencing mechanism of urban flooding. Exploring the evolution of the mechanisms influencing urban flooding in the Guangdong Hong Kong Macao Greater Bay Area (GBA) urban agglomeration is of great reference significance for formulating flood prevention and control measures and promoting high-quality development of the GBA city cluster. In this paper, we fully use the collected information on urban flooding events from 1980 to 2018 in the GBA city cluster. Correlation analysis and geographically weighted regression (GWR) are used to analyze the influence of impervious surface percentage (ISP), impervious surface aggregation index (AI), impervious surface mean shape index (Shape_MN), vegetation cover (FVC), water surface ratio (WSR), relative elevation (RE) and slope on flooding in urban clusters and their evolution characteristics over time from a global perspective and spatial heterogeneity, respectively. The results show that: 1) ISP, AI, Shape_MN, and WSR are positively correlated with urban flooding, while FVC, RE, and Slope are negatively correlated with urban flooding. The correlations of each factor showed a general trend of gradual strengthening over time, and the increase rate slowed down after 2000, while the correlation of WSR showed a relatively noticeable decrease. 2) The GWR results show that each factor’s influence on urban flooding has pronounced spatial-temporal heterogeneity, and each factor shows different distribution characteristics. This study uses long time series of urban flooding point data to explore the spatial-temporal evolution of the influencing mechanism of urban flooding in the GBA urban agglomeration. We hope to provide a scientific basis for an in-depth understanding of the causes of urban flooding in the GBA, intending to provide auxiliary decision-making support for the formulation of waterlogging prevention and control measures.

In recent years, object detection has shown excellent results on a large number of annotated data, but when there is a discrepancy between the annotated data and the real test data, the performance of the trained object detection model is often degraded when it is directly transferred to the real test dataset. Compared with natural images, remote sensing images have great differences in appearance and quality. Traditional methods need to re-label all image data before interpretation, which will consume a lot of manpower and time. Therefore, it is of practical significance to study the Cross-Domain Adaptation Object Detection (CDAOD) of remote sensing images. To solve the above problems, our paper proposes a Rotation-Invariant and Relation-Aware (RIRA) CDAOD network. We trained the network at the image-level and the prototype-level based on a relation aware graph to align the feature distribution and added the rotation-invariant regularizer to deal with the rotation diversity. The Faster R-CNN network was adopted as the backbone framework of the network. We conducted experiments on two typical remote sensing building detection datasets, and set three domain adaptation scenarios: WHU 2012 → WHU 2016, Inria (Chicago) → Inria (Austin), and WHU 2012 → Inria (Austin). The results show that our method can effectively improve the detection effect in the target domain, and outperform competing methods by obtaining optimal results in all three scenarios.

Semantic segmentation from very fine resolution (VFR) urban scene images plays a significant role in several application scenarios including autonomous driving, land cover classification, and urban planning, etc. However, the tremendous details contained in the VFR image, especially the considerable variations in scale and appearance of objects, severely limit the potential of the existing deep learning approaches. Addressing such issues represents a promising research field in the remote sensing community, which paves the way for scene-level landscape pattern analysis and decision making. In this paper, we propose a Bilateral Awareness Network which contains a dependency path and a texture path to fully capture the long-range relationships and fine-grained details in VFR images. Specifically, the dependency path is conducted based on the ResT, a novel Transformer backbone with memory-efficient multi-head self-attention, while the texture path is built on the stacked convolution operation. Besides, using the linear attention mechanism, a feature aggregation module is designed to effectively fuse the dependency features and texture features. Extensive experiments conducted on the three large-scale urban scene image segmentation datasets, i.e., ISPRS Vaihingen dataset, ISPRS Potsdam dataset, and UAVid dataset, demonstrate the effectiveness of our BANet. Specifically, a 64.6% mIoU is achieved on the UAVid dataset. Code is available at https://github.com/WangLibo1995/GeoSeg.