Abstract: Riparian vegetation coverage is widely recognized as a crucial indicator for evaluating the health and functioning of river and lake ecosystems. It performs essential ecological functions, including stabilizing riverbanks, regulating microclimates, filtering pollutants, conserving water resources, and providing habitats for diverse species. Consequently, developing efficient and cost-effective methods to monitor riparian vegetation coverage while improving measurement precision has become a pressing scientific challenge in environmental monitoring and ecosystem management. Traditional field investigation methods, while providing accurate ground-truth data, face notable limitations such as high time costs, intensive labor requirements, and restricted spatial coverage. These constraints make them impractical for large-scale or frequent monitoring efforts. In recent years, unmanned aerial vehicle (UAV) remote sensing technology has attracted considerable attention due to its operational flexibility, high spatial resolution, cost-effectiveness, and ability to function under various weather conditions. This advancement provides innovative solutions for rapid, detailed, and repeatable vegetation assessment in riparian environments. This study developed and validated a classification methodology based on vegetation index thresholds derived from UAV remote sensing data. The research employed two vegetation indices: the widely used Normalized Difference Vegetation Index (NDVI) derived from multispectral data and the Visible-band Difference Vegetation Index (VDVI) calculated from visible spectrum data. To determine optimal classification thresholds, three automated thresholding algorithms were implemented and compared: the OTSU algorithm (Otsu's method), which maximizes between-class variance; the histogram entropy method, which applies information theory concepts; and the bimodal histogram method, which identifies the optimal separation between vegetation and non-vegetation classes. The methodology was applied to a representative section of the Jurong River in Nanjing, China, where vegetation coverage was quantitatively calculated. A comprehensive accuracy assessment was conducted to evaluate classification performance and confirm the feasibility of the proposed approach. The results demonstrated that both NDVI, derived from multispectral data, and VDVI, obtained from visible-band data, are effective and reliable indicators for feature classification and vegetation coverage estimation in riparian zones. When NDVI was used as the classification basis, all three thresholding methods (OTSU, histogram entropy, and bimodal histogram) exhibited excellent performance across diverse riparian environments. The three methods achieved remarkable consistency, with overall classification accuracy exceeding 93% and Kappa coefficients above 0.806. These results indicate strong agreement between classification outcomes and field validation data, confirming the stability and applicability of NDVI for accurately calculating riparian vegetation coverage. When VDVI was employed as the primary classification indicator, the bimodal histogram threshold method proved most effective for determining optimal thresholds, particularly in areas with dense vegetation coverage. In cases where the VDVI grayscale histogram failed to exhibit a clear bimodal distribution, the OTSU algorithm provided a viable alternative. Although this substitution produced slightly lower accuracy, it remained practically useful for vegetation monitoring. Under high-coverage conditions, combining cost-effective visible-light UAVs for VDVI data acquisition with the bimodal histogram method for coverage estimation proved especially effective. This approach sustained acceptable accuracy while significantly reducing costs, offering a practical solution for large-scale monitoring under budget constraints. This study establishes a comprehensive, efficient, and transferable methodological framework for riparian vegetation coverage estimation that successfully integrates theoretical sophistication with practical applicability. The framework effectively balances the dual demands of monitoring efficiency and measurement precision while offering significant advantages over conventional approaches. The proposed methodology not only reduces the high costs associated with traditional manual surveys but also provides reliable technical support for refined ecosystem management and dynamic environmental monitoring in riparian zones. It demonstrates considerable potential for application in large-scale ecological assessment programs and recurring monitoring initiatives where both accuracy and economic feasibility are paramount considerations.