Deriving structural and biochemical crop traits from one UAV sensor: Investigating a multiband VNIR/SWIR imaging system for crop trait monitoring

Abstract. Frame-based VNIR/SWIR multispectral sensors on UAVs offer promising capabilities for precision agriculture by enabling the easy simultaneous acquisition of spectral and structural crop information. This study provides an independent validation of a two-band VNIR/SWIR sensor system for monitoring winter wheat traits and compares the results with previous findings. The UAV flights were conducted on a single date (May 11, 2022), capturing image datasets at wavelengths of 910, 980, 1100, 1200, 1510, and 1650 nm. Structure from Motion (SfM) processing enabled crop height extraction from the same multispectral datasets. Ground-truth data included fresh and dry biomass, moisture, nitrogen concentration, and nitrogen uptake from 36 samples across six varieties and three fertilization levels. Bivariate regression analyses revealed moderate performance for spectral vegetation indices (NRI: R²=0.52–0.61; GnyLi: R²=0.50–0.62), which was lower than that previously reported. Crop height showed a superior predictive capability (R²=0.63–0.75), demonstrating consistency across studies. Multivariate models combining vegetation indices with crop height significantly improved trait estimation (R²=0.72–0.84, nRMSE=0.12–0.15), confirming that integrated spectral-structural approaches provide robust performance even when individual predictors show limitations. While this single-date analysis limits conclusions about temporal stability throughout the growing season, it provides valuable validation of the capabilities of the sensor system. The ability to derive both structural and biochemical data from single-sensor imagery is the key advantage of this camera system. Future research should expand to multi-temporal analyses across complete growing seasons and implement the recently developed 6-channel VNIR/SWIR system to address the current limitations. This study reinforces the fact that combining SWIR spectral features with structural parameters is essential for reliable estimation of crop traits.