Accurate building footprint data is indispensable for urban planning, and database creation but, existing building extraction models, primarily originating from developed countries, face limitations when applied to rural regions or developing countries due to intricate building patterns across diverse land-use types. This research addresses these challenges, focusing on a case study of Bangkok province, Thailand. We introduce a comprehensive methodology that develops seven specialized YOLOv8 and YOLOv7 models, trained from scratch to accommodate distinct land-use types and their unique building patterns. A multi-domain assessment is conducted across various land-use types to assess the accuracy, applicability, and adaptability of these models. The findings offer a valuable template for future research in building extraction from satellite imagery, advancing instance segmentation techniques, and enhancing their practical utility across diverse land-use scenarios.

Remote sensing image semantic segmentation of land use, benefitted from the development of deep learning and consequently made considerable progress in terms of inferencing accuracy and speed. However, the effective training of semantic segmentation models for remote sensing imagery necessitates extensively detailed pixel-level annotations, and gathering such data is both time-intensive and laborious. Thus, this study implemented low-rank adaptation on a stable diffusion algorithm to learn the distribution of the pixel-level annotations in case of the LoveDA dataset. Consequently, the annotation-image pairs were used to train the remote sensing image generator based on stable diffusion guided by ControlNet. We proposed a stable diffusion based approach, which can generate image-annotation pairs from scratch. The generated annotation and image pairs achieved a high accuracy of 0.520 mean intersection-over-union on LoveDA dataset, which is close to the original data training result of 0.539 mIoU. Furthermore, the mixed training using generated and original data achieved 0.542 mIoU, thereby demonstrating the data augmentation function of our approach. This study provided a solution for the high-cost pixel-level annotation issue, and thus, exhibited the potential of artificial intelligence generated content.

Land price is an important economic factor in producing meaningful references for regional planners by assisting them in urban planning, economic decision-making, and land resource allocation. However, related studies in land price analysis were mainly focused on the factors of site area and plot ratio, analysis of the potential impact of streetscape factors and human subjective perception on the land price has been lacking, regardless of the impact on supply and demand relationship. Therefore, this study developed a new approach for estimating and analyzing land prices through deep learning that considered the streetscape and human subjective perception factors. In the estimation part, we developed a fine-grained end-to-end deep learning model, the input is street view images, and the output is land prices. In the analysis part, we extracted the semantic segmentation results and human subjective perception scores from the images and combined them with the results of land price estimation. We then introduced a combination of quantitative analysis using the gradient-weighted class activation mapping and L1-based sparse linear regression to model the relationship between the streetscape and the human subjective perception quantitatively. The gradient-weighted class activation mapping was used to determine which categories of pixels deep learning relied on to output the results of land price estimation quantitatively. Combined with segmentation results, we implemented L1-based sparse linear regression and quantitatively determined the importance of the streetscape factors for land prices. Overall, our deep learning model achieved 77.99 % accuracy by only using street views in estimating the land price, and we illustrated the impacts of the streetscape and perception on the land price by showing that perception scores are more important than streetscape factors such as road and mountain in the streetscape. Comfortable, lived-in feel in perception have the most important impact on land price estimation.

Building footprints are one of the primary data sources in the urban geographic information system (GIS) database. They are crucial for applications such as urban planning and population estimation. However, high-quality remote sensing imagery is not always available as open-sourced data for a wide area. Therefore, a versatile method was proposed in this study for building footprint extraction pipelines for a wide area. The proposed method was based on an instance segmentation approach using open-sourced satellite images. First, a reference-based color normalization algorithm was applied to eliminate the color differences in the satellite images from different sources. Subsequently, a mask R-CNN model was fine-tuned for the footprint extraction task on Detectron2. Finally, a modified image mosaicking approach was proposed for a large scene. The Hyogo prefecture, Japan, was used as the test area to evaluate the performance of the method.

People flow data are important for economic prediction and urban planning development. In this study, we deployed a Swin Transformer on panoramic street view images and constructed a wide-area people flow trend model using image data. First, for people flow trend prediction, we concentrated the flow data of a wide area and counted it into mesh units, and defined people flow level by the number of people on each mesh unit. Second, to construct the dataset, we determined which mesh unit each image belonged to by the locations of photography and coordinated range of mesh units to define people flow trend level for each image. Then, after category balance and random selection from wide-area data, we implemented Swin Transformer and trained models in this small part of the data, and made the precision evaluation and output a 75.84% accuracy model. Finally, we deployed our model in the total study area data and contributed to the people flow trend model.