Automatic road extraction from very-high-resolution remote sensing images has become a popular topic in a wide range of fields. Convolutional neural networks are often used for this purpose. However, many network models do not achieve satisfactory extraction results because of the elongated nature and varying sizes of roads in images. To improve the accuracy of road extraction, this paper proposes a deep learning model based on the structure of Deeplab v3. It incorporates squeeze-and-excitation (SE) module to apply weights to different feature channels, and performs multi-scale upsampling to preserve and fuse shallow and deep information. To solve the problems associated with unbalanced road samples in images, different loss functions and backbone network modules are tested in the model’s training process. Compared with cross entropy, dice loss can improve the performance of the model during training and prediction. The SE module is superior to ResNext and ResNet in improving the integrity of the extracted roads. Experimental results obtained using the Massachusetts Roads Dataset show that the proposed model (Nested SE-Deeplab) improves F1-Score by 2.4% and Intersection over Union by 2.0% compared with FC-DenseNet. The proposed model also achieves better segmentation accuracy in road extraction compared with other mainstream deep-learning models including Deeplab v3, SegNet, and UNet.

Nighttime light (NTL) data is recognized as a reliable proxy for measuring the scope and intensity of human activity, finding wide application in studies such as urbanization monitoring, socioeconomic estimation, and ecological environment assessment. However, the substantial discrepancies and limited temporal coverage of existing NTL datasets have constrained their potential for long-term research applications. To address this, a Nighttime Light U-Net super-resolution network is proposed for the cross-sensor calibration between the Defense Meteorological Satellite Program-Operational Linescan System (DMSP-OLS) NTL data and the Suomi National Polar-Orbiting Partnership-Visible Infrared Imaging Radiometer Suite (NPP-VIIRS) NTL data. This network is applied to generate a continuous and consistent 500-meter global annual simulated VIIRS NTL dataset (SVNL) from 1992 to 2023. Validation results indicate a high confidence in the quality of the SVNL data, demonstrating its superiority in capturing longer NTL dynamics, maintaining higher temporal consistency, and presenting greater spatial detail compared with other NTL datasets. The SVNL could be utilized for prolonged human activities monitoring, and further research on regional or global urbanization.

The worldwide development of multi-center structures in large cities is a prevailing development trend. In recent years, China’s large cities developed from a predominantly mono-centric to a multi-center urban space structure. However, the definition and identification city centers is complex. Both nighttime light data and point of interest (POI) data are important data sources for urban spatial structure research, but there are few integrated applications for these two kinds of data. In this study, visible infrared imaging radiometer suite (NPP-VIIRS) nighttime imagery and POI data were combined to identify the city centers in Hangzhou, China. First, the optimal parameters of multi-resolution segmentation were determined by experiments. The POI density was then calculated with the segmentation results as the statistical unit. High–high clustering units were then defined as the main centers by calculating the Anselin Local Moran’s I, and a geographically weighted regression model was used to identify the subcenters according to the square root of the POI density and the distances between the units and the city center. Finally, a comparison experiment was conducted between the proposed method and the relative cut-off_threshold method, and the experiment results were compared with the evaluation report of the master plan. The results showed that the optimal segmentation parameters combination was 0.1 shape and 0.5 compactness factors. Two main city centers and ten subcenters were detected. Comparison with the evaluation report of the master plan indicated that the combination of nighttime light data and POI data could identify the urban centers accurately. Combined with the characteristics of the two kinds of data, the spatial structure of the city could be characterized properly. This study provided a new perspective for the study of the spatial structure of polycentric cities.