Assessing the Accuracy of UAV Photogrammetry for Urban Land Use Mapping

Thesis Abstract

Accurate land use mapping in urban environments is essential for sustainable urban planning and effective resource management, yet challenges persist in achieving precise and reliable data, particularly when utilizing remote sensing technologies. This study aims to evaluate the accuracy of unmanned aerial vehicle (UAV) photogrammetry for urban land use classification, focusing on its potential to serve as a cost-effective and rapid alternative to traditional survey methods. The specific objectives include assessing spatial accuracy, evaluating classification correctness, and investigating the influence of UAV flight parameters on data quality. Employing an explanatory sequential mixed-methods research design, the study integrates quantitative and qualitative approaches to comprehensively evaluate UAV-based imagery. The population comprises urban areas within the metropolitan city of Metroville, with a sample size of 25 land parcels encompassing diverse land use types such as residential, commercial, industrial, and green spaces. A stratified random sampling technique ensures representative coverage across land use categories. Data collection involved deploying a DJI Phantom 4 Pro UAV equipped with a high-resolution multispectral camera to capture aerial imagery at varying altitudes (50 m, 100 m, and 150 m) across each selected site. Ground truth data were collected via GPS-accurate field surveys and high-resolution terrestrial photographs. The study also gathered metadata including UAV flight parameters, ambient weather conditions, and image acquisition settings. Image processing and processing involved using Pix4Dmapper software to generate orthomosaics, digital surface models, and 3D point clouds. Land use classification was performed through object-based image analysis (OBIA) utilizing eCognition software, with classification accuracy assessed via confusion matrices, overall accuracy, and the Kappa coefficient. Spatial accuracy was evaluated by comparing UAV-derived data with ground truth references through root mean square error (RMSE) analysis. To examine the relationships among UAV flight parameters, image quality, and classification accuracy, multiple linear regression analyses were conducted. The potential influence of image resolution, flight altitude, and overlap percentage on mapping accuracy was a key focus. The anticipated results include high classification accuracy with overall accuracies exceeding 85% and Kappa coefficients above 0.80 at optimal flight altitude (50 m), with decreasing accuracy observed at higher altitudes. Spatial accuracy metrics are expected to reveal RMSE values below 0.5 meters for features larger than 5 meters, demonstrating the method’s suitability for detailed urban land use mapping. Regression analysis is projected to indicate a significant positive correlation between lower flight altitude and higher spatial and classification accuracy, moderated by environmental conditions and camera settings. This research contributes to the field by providing empirical evidence of UAV photogrammetry’s precision and reliability in complex urban landscapes, thereby filling current gaps regarding optimal data acquisition parameters for urban land use classification. It advances the theoretical understanding of remote sensing accuracy, drawing on Tobler’s First Law of Geography and the Theory of Spatial Data Quality, to interpret spatial data reliability. The findings will inform urban planners, GIS practitioners, and policymakers regarding the deployment thresholds and limitations of UAV-based land use mapping, emphasizing cost-effectiveness and operational efficiency. The study concludes that UAV photogrammetry, when conducted within optimal flight parameters, offers a highly accurate, timely, and economical approach to urban land use mapping. Recommendations include adopting lower flight altitudes for detailed mapping, integrating UAV data with existing GIS datasets for enhanced spatial analysis, and developing standardized protocols for UAV-based urban land surveys. Future research should explore the integration of multispectral and hyperspectral sensors, as well as advanced machine learning algorithms for automated classification, to further enhance mapping accuracy and operational feasibility in diverse urban contexts.