Commercial Off-the-Shelf Digital Cameras on Unmanned Aerial Vehicles for Multitemporal Monitoring of Vegetation Reflectance and NDVI

Abstract

This paper demonstrates the ability to generate quantitative remote sensing products by means of an unmanned aerial vehicle (UAV) equipped with one unaltered and one near infrared-modified commercial off-the-shelf (COTS) camera. Radiometrically calibrated orthomosaics were generated for 17 dates, from which digital numbers were corrected to surface reflectance and to normalized difference vegetation index (NDVI). Validation against ground measurements showed that 84%–90% of the variation in the ground reflectance and 95%–96% of the variation in the ground NDVI could be explained by the UAV-retrieved reflectance and NDVI, respectively. Comparisons against Landsat 8 data showed relationships of $0.73\leq R^{2} \geq 0.84$ for reflectance and $0.86\leq R^{2} \geq 0.89$ for NDVI. It was not possible to generate a fully consistent time series of reflectance, due to variable illumination conditions during acquisition on some dates. However, the calculation of NDVI resulted in a more stable UAV time series, which was consistent with a Landsat series of NDVI extracted over a deciduous and evergreen woodland. The results confirm that COTS cameras, following calibration, can yield accurate reflectance estimates (under stable within-flight illumination conditions), and that consistent NDVI time series can be acquired in very variable illumination conditions. Such methods have significant potential in providing flexible, low-cost approaches to vegetation monitoring at fine spatial resolution and for user-controlled revisit periods.