Photovoltaic (PV) power generation has become a key strategy in Taiwan's transition to green energy, with Tainan City repeatedly ranking as the leading municipality in PV electricity generation. However, the impact of solar panels on surrounding land surface temperature (LST) remains highly controversial. This study employs remote sensing techniques to examine the influence of PV installations on LST. By analyzing temperature variations across different land use types and time periods, the research aims to assess the thermal effects associated with PV deployment. The results reveal that LST decreases with increasing distance from PV panels, with the most significant temperature drop occurring approximately 25 meters away. Additionally, the conversion of vegetated or aquatic areas into bare land or PV-covered land tends to elevate in surface temperature. In some regions, bare land exhibited a temperature increase of approximately 1 to 2?°C compared to PV-covered surfaces during the same period.

In recent years, point clouds have become indispensable data for producing multi-dimensional spatial information. However, "Quality" has long been overlooked in point cloud surveying and mapping workflows because, in practical applications, it is often treated as an end-stage validation or an auxiliary indicator for evaluating results, with little attention to its influence on operational decisions, strategic planning, and the adoption of survey products. This paper aims to demonstrate the critical importance of quality for advancing point cloud surveying technology, arguing that quality not only governs the credibility of survey outputs but also fundamentally shapes the entire surveying workflow; through multifaceted examinations, how quality forms risks, constraints, or decision-making references at different levels is thoroughly elaborated; and most importantly, a systematic conceptual framework for quality awareness with a clear problem consciousness and research foundation for the subsequent development of quality-oriented point cloud surveying workflows and decision-support methodologies has been established.

This study integrates Unmanned Aerial Vehicles (UAVs) with the deep learning object detection model YOLOv9 to conduct automated detection and analysis of cracks in asphalt pavement. Leveraging the high mobility and high-resolution imaging capabilities of UAVs, the system can rapidly cover large road areas and significantly reduce the time and labor costs associated with manual inspections. The YOLOv9 model was initially trained using the publicly available CrackForest dataset for pavement cracks. After 100 training epochs, the model achieved a detection accuracy of mAP50 = 0.891 and mAP50-95 = 0.550 on the test set, demonstrating strong performance in small object detection and pavement crack identification. Finally, through orthorectification and geometric correction, the detection results were projected onto real-world coordinates for crack width quantification. The results verify that the proposed system can serve as an effective tool for road maintenance decision-making and management.

This study proposes an automated 3D modeling workflow for large cylindrical salt warehouses by integrating photogrammetry and deep learning techniques, aiming to improve modeling efficiency and measurement accuracy under harsh environmental conditions. Two industrial cameras are mounted on a central overhead crane to capture high-overlap images, which are enhanced using Contrast Limited Adaptive Histogram Equalization. Ground control point targets are detected using YOLOv8, followed by subpixel localization through Normalized Cross-Correlation. To improve the quality of aerial triangulation, LightGlue is employed for feature matching and redundant image filtering. A Metashape Python script is then used to automate bundle adjustment, error evaluation, and dense point cloud generation. Experiments conducted with 401 GCP samples show that the positioning error is approximately 2%, demonstrating the stability and accuracy of the proposed approach. The results indicate that the method effectively reduces the need for manual field measurements and improves the efficiency of salt pile volume estimation.