Unmanned aerial vehicles (UAVs) have emerged as key platforms in remote sensing and precision agriculture owing to their high spatial resolution, relatively low cost, and ability to acquire imagery at flexible time intervals. This study evaluates the feasibility of using multispectral imagery collected by a UAV-mounted Sequoia sensor to estimate biophysical and biochemical parameters of rice. Data were acquired at three phenological stages over paddy fields at the CAPIC Agricultural Research Centre, Mazandaran Province, Iran. In the first stage, UAV images underwent quality assessment and denoising, followed by radiometric and geometric corrections. Selected spectral indices—NDVI, GNDVI, LCI, NDRE, NDWI, and SIPI2—were then derived to provide information on canopy greenness, chlorophyll content, leaf water status, and overall physiological condition. Subsequently, vegetation canopy reflectance was simulated using the physically based PROSAIL model, and an inversion via look-up tables (LUTs) was applied to retrieve key biophysical parameters, including leaf area index (LAI), leaf chlorophyll content (Cab), leaf equivalent water thickness (Cw), and the leaf structure parameter (N). Field measurements were used for validation, and model performance was assessed using the coefficient of determination (R²) and the root mean square error (RMSE). Results indicate that UAV multispectral imagery enables accurate retrieval of plant parameters and that PROSAIL effectively captured the phenological dynamics of rice. In particular, estimates of chlorophyll-related metrics and leaf equivalent water thickness exhibited very high accuracy (R² ≈ 0.996), with only minor errors observed at certain growth stages (RMSE < 0.2). Overall, integrating UAV multispectral data with physically based models such as PROSAIL offers an effective tool for crop growth monitoring and the optimisation of input management within precision agriculture. Nonetheless, limitations such as restricted spatial coverage and the relatively high cost of imaging systems remain; future work could mitigate these by expanding spatial extent and leveraging machine-learning and deep-learning algorithms.