Detection of homogeneous wheat areas using multi-temporal UAS images and ground truth data analyzed by cluster analysis

Abstract

Vegetation indices (VIs) obtained from unmanned aerial system (UAS) are effective for monitoring quantitative and qualitative characteristics of vegetation cover. Nevertheless, the identification of agronomic homogeneous crop areas to be managed in a specific different agronomic way is still to be improved in precision farming. The aim of the study was to reduce information gap on the detection of homogeneous wheat areas using multi-temporal remote sensing image and agronomic crop traits by cluster analysis. The images were acquired by an eBee UAS on a small plot of 720 m2 at three different crop growth stages High-resolution orthoimages (3.5 cm·pixel−1) were generated by Pix4D and QGIS. At each growth stage, biometric ground-truth data and VIs (NDVI and SAVI) were clustered to detect homogeneous crop areas. At tillering and anthesis stages, three significant homogeneous areas with low (L) medium (M) and high (H) VIs and agronomical values were identified for both indices. Yield-related traits (at harvest) and VIs (at anthesis) confirmed that L and M areas, with agronomic constraints identified at anthesis, showed crop yield losses at harvest. Cluster analysis, using UAS and ground truth data, has proved to be a good strategy to identify the homogeneous wheat crop areas.

Keywords

• Durum wheat
• agronomic traits
• vegetation indices
• UAV
• precision agriculture
• cluster analysis