A5021485937- Remote-Sensing Image Classification
- Remote Sensing in Agriculture
- Remote Sensing and Land Use
- Advanced Image and Video Retrieval Techniques
- Asphalt Pavement Performance Evaluation
- Automated Road and Building Extraction
- Non-Destructive Testing Techniques
- Atmospheric chemistry and aerosols
- Air Quality and Health Impacts
- Air Quality Monitoring and Forecasting
- Computational Geometry and Mesh Generation
- 3D Shape Modeling and Analysis
- Advanced Image Fusion Techniques
- Infrastructure Maintenance and Monitoring
- Land Use and Ecosystem Services
Shandong University of Technology
2022-2024
Nanjing University of Information Science and Technology
2024
The semantic segmentation of land use and cover (LULC) is a crucial widely employed remote sensing task. Conventional convolutional neural networks vision transformers have been extensively utilized for LULC segmentation. However, high-resolution images contain wealth spatial color texture information, which not fully exploited by traditional deep learning approaches. information bottleneck CNNs results in the loss significant amount detail during feature extraction process, further limits...
As core algorithms of geographic computing, overlay analysis typically have computation-intensive and data-intensive characteristics. It is highly important to optimize by parallelizing the vector polygons after reasonable data division. To address problem unbalanced partitioning in task decomposition process for parallel polygon calculation, this paper presents a method based on shape complexity index optimization, which achieves equalization among multicore computing tasks. Taking...
Building change detection is an important task in the remote sensing field, and powerful feature extraction ability of deep neural network model shows strong advantages this task. However, datasets used for study are mostly three-band high-resolution images from a single data source, few spectral features limit development building multisource images. To investigate influence texture on effect based learning, dataset (MS-HS BCD dataset) produced paper using GF-1 Sentinel-2B multispectral...
Road crack segmentation based on high-resolution images is an important task in road service maintenance. The undamaged surface area much larger than the damaged a highway. This imbalanced situation yields poor performance for convolutional neural networks. In this paper, we first evaluate mainstream network structure task. Second, inspired by second law of thermodynamics, improved method called recurrent adaptive pixelwise proposed to solve extreme imbalance between positive and negative...
Deep convolutional networks often encounter information bottlenecks when extracting land object features, resulting in critical geometric loss, which impedes semantic segmentation capabilities complex geospatial backgrounds. We developed LULC-SegNet, a network for use and cover (LULC), integrates features from the denoising diffusion probabilistic model (DDPM). This enhances clarity of edge segmentation, detail resolution, visualization accuracy contours by delving into spatial details...
Automatic feature semantic segmentation of remote sensing images is an extremely critical research direction in the field geographic information science. Especially vast and complex desert area, wide spatial distribution surface features, texture characteristics uneven sample classification bring great challenges to recognition features. In response question, we propose innovative network scheme, which a that combines dynamic convolutional decomposition extraction multi-scale deformable...
This study examined the impact of temporary air quality control measures on reducing pollutants during 2022 Winter Olympics in China, utilizing real-time monitoring data from 2017 and to assess spatial temporal variations critical pollutant concentrations. The results showed that concentrations PM2.5, PM10, CO, SO2, NO2 Beijing–Tianjin–Hebei region Olympic Games a marked decrease compared historical period, with reductions 36.59%, 20.35%, 33.95%, 28.90%, 22.70%, respectively. Significant...