ASABE Technical Library - Abstract

Click on “Download PDF” for the PDF version or on the title for the HTML version.
If you are not an ASABE member or if your employer has not arranged for access to the full-text, view access and subscription options

Cabbage volume estimation and water stress monitoring using combined UAV and ground-based multispectral and thermal imaging approaches

Published by the American Society of Agricultural and Biological Engineers, St. Joseph, Michigan www.asabe.org

Citation:  2025 ASABE Annual International Meeting  2500876.(doi:10.13031/aim.202500876)
Authors:   Md Nasim Reza, Kyu-Ho Lee, Sun-Ok Chung
Keywords:   Precision agriculture, Cabbage, UAV imaging, Data fusion, Machine learning, Volume estimation

Abstract. Accurate estimation of crop volume and effective water stress detection are critical for enhancing agricultural productivity and optimizing water resource management. This study investigated an integrated approach for cabbage (Brassica rapa subsp. pekinensis) volume estimation and water stress monitoring, utilizing unmanned aerial vehicle (UAV) and ground-based multispectral and thermal imaging platforms. A commercial UAV with RTK positioning, equipped with multispectral and thermal sensors, captured aerial imagery at a 30 m altitude, while a ground-based imaging platform at 1.8 m collected high-resolution data. Preprocessing included radiometric calibration, orthomosaic generation, georeferencing, and normalization to ensure compatibility between UAV and ground datasets. Vegetation indices such as NDVI and GNDVI were calculated to assess plant vigor and water stress, while canopy temperature was used to derive the Crop Water Stress Index (CWSI). For cabbage volume estimation, a deep learning-based segmentation model was employed to detect individual plants from multispectral images. The model segmented cabbage heads, extracted morphological features, and estimated volume using a regression approach. NDVI and GNDVI showed strong negative linear correlations with the CWSI, with R² values of 0.9412 and 0.9249, respectively, showed the reliability in detecting crop water stress. Additionally, UAV-derived CWSI demonstrated a strong correlation (R² = 0.8323) with field-measured values, validating the feasibility of UAV thermal imaging for spatially explicit stress detection. Spatial NDVI and thermal imagery further illustrated canopy variability across the field. For cabbage volume estimation, a U-Net segmentation model trained on annotated data accurately identified individual cabbage heads. Area and volume estimations based on image pixel counts were validated against ground and actual measurements, yielding high correlations (R² > 0.86). This approach supports precision agriculture through targeted irrigation and resource use. Future work will refine stress mapping, enable real-time decisions, and broaden crop applicability.

(Download PDF)    (Export to EndNotes)