This study utilizes Thermal Infrared (TIR) imaging technology to detect hotspots in photovoltaic (PV) modules of solar power plants. Unmanned aerial vehicle (UAV)-based TIR imagery is crucial for efficiently analyzing fault detection in solar power plants. This research explores optimal operational parameters for generating high-quality TIR images using UAV technology. In addition to existing variables such as humidity, emissivity, height, wind speed, irradiance, and ambient temperature, newly considered variables including the angle of incidence between the target object and the thermal infrared camera are analyzed for their impact on TIR images. Based on the solar power plant’s tilt (20°) and the location coordinate data of the hotspot modules, the inner and outer products of the vectors were used to obtain the normal vector and angle of incidence of the solar power plant. It was discovered that the difference between measured TIR temperature data and Land Surface Temperature (LST) data varies with changes in the angle of incidence. The analysis presented in this study was conducted using multiple regression analysis to explore the relationships between dependent and independent variables. The Ordinary Least Squares (OLS) regression model employed was able to explain 63.6% of the variability in the dependent variable. Further, the use of the Condition Number (Cond. No.) and the Variance Inflation Factor (VIF) revealed that the multicollinearity among all variables was below 10, ensuring that the independence among variables was well-preserved while maintaining statistically significant correlations. Furthermore, a positive correlation was observed with the actual measured temperature values, while a negative correlation was observed between the TIR image data values and the angle of incidence. Moreover, it was found that an angle of incidence between 15° and 20° yields the closest similarity to LST temperature data. In conclusion, our research emphasizes the importance of adjusting the angle of incidence to 15–20° to enhance the accuracy of TIR imaging by mitigating overestimated TIR temperature values.

Earthwork volume calculation is a fundamental process in civil engineering and construction, requiring high-precision terrain data to assess ground stability encompassing load-bearing capacity, susceptibility to settlement, and slope stability and to ensure accurate cost estimation. However, seasonal and environmental constraints pose significant challenges to surveying. This study employed unmanned aerial vehicle (UAV) photogrammetry and light detection and ranging (LiDAR) mapping to evaluate the accuracy of digital terrain model (DTM) generation and earthwork volume estimation in densely vegetated areas. For ground extraction, color-based indices (excess green minus red (ExGR), visible atmospherically resistant index (VARI), green-red vegetation index (GRVI)), a geometry-based algorithm (Lasground (new)) and their combinations were compared and analyzed. The results indicated that combining a color index with Lasground (new) outperformed the use of single techniques in both photogrammetric and LiDAR-based surveying. Specifically, the ExGR–Lasground (new) combination produced the most accurate DTM and achieved the highest precision in earthwork volume estimation. The LiDAR-based results exhibited an error of only 0.3% compared with the reference value, while the photogrammetric results also showed only a slight deviation, suggesting their potential as a practical alternative even under dense summer vegetation. Therefore, although prioritizing LiDAR in practice is advisable, this study demonstrates that UAV photogrammetry can serve as an efficient supplementary tool when cost or operational constraints are present.

This study examines how train–test ground sampling distance (GSD) mismatch affects UAV-based oriented vehicle detection under controlled source-scene and network-input conditions. Five datasets at 10, 15, 20, 25, and 30 cm/pixel were generated by direct downsampling from the same 1509 UAV source frames while preserving vehicle annotations and data-partition assignments. The GSD conversion procedure was empirically evaluated using 25 image measurements, yielding an overall mean absolute error of 0.180 cm/pixel, root mean square error of 0.225 cm/pixel, and mean absolute relative error of 0.891%. YOLO11m-OBB models were trained under two augmentation settings and three random seeds. The best validation checkpoint from each of 30 training runs was evaluated across all five test GSDs, producing 150 primary cross-GSD evaluations. Mean matched and mismatched ${AP}_{50-95}$ were 0.7749 and 0.4915 under controlled augmentation and 0.8166 and 0.7346 under default augmentation, respectively. Across the five equally weighted test GSDs, 15 cm/pixel training achieved the highest mean ${AP}_{50-95}$ reaching 0.6249 ± 0.0026 and 0.8049 ± 0.0018 under controlled and default augmentation. For 15-to-30 cm/pixel transfer, recall at a confidence of 0.25 increased from 0.5967 to 0.8407 with default augmentation. Nevertheless, recall remained only 0.0495 for vehicles with an OBB short side below 2 pixels under matched 30 cm/pixel evaluation. These results show that cross-GSD performance depends on augmentation, relative sampling, and object pixel footprint. The findings are limited to the UAV-resampling conditions evaluated here and do not establish a universal optimal GSD or direct transfer to actual satellite imagery.