Share | 04/24/2020
Abstract: This study aims to develop prediction models for Brix, Pol, fiber, and CCS value in cane fields by using vegetation indices calculated from canopy reflectance based on images taken by a multispectral camera mounted on a small unmanned aerial system (UAS). For this purpose, a MicaSense RedEdge equipped with six-rotor UAS was used. This experiment was conducted on three sugarcane varieties, namely UT84-12 (a flood-tolerant variety), and K88-92 and KK3 (drought-tolerant varieties). The acquired images were generated into five bands of reflectance maps (blue, green, red, NIR, and rededge) in Pix4D software, which were used to produce a map of six vegetation indices (GNDVI, NDVI, RVI, CIgreen, CIrededge, and SRPIb). After that, each vegetation index value of the cropped plot was averaged. After the flight mission, average values of Brix, Pol, CCS, and fiber were measured from four randomly chosen sugarcane stalks from each variety. The six vegetation indices were paired with averaged references values to create simple linear regression models. It was found that CIrededge shows the best correlation with Pol (R² 0.77) and CCS (R² 0.68), independent of variety, while Brix models are dependent on the variety and require different vegetation indices. For instance, the Brix values of drought-tolerant K88-92 and KK3 correlate best with CIrededge, which is most sensitive to chlorophyll in canopy (R² 0.91), while the flood-tolerant variety (UT84-12) shows good correlation with SRPIb, sensitive to nitrogen leaf (R² 0.87). We found a poor correlation (R² = 0.35–0.50) between fiber content with all six vegetation indices for all varieties.
Read the paper
Authors: Chea, Chanreaksa & Saengprachatanarug, Khwantri & Posom, Jetsada & Wongphati, Mahisorn & Taira, Eizo
Associations: Khon Kaen University, King Mongkut’s Institute of Technology Ladkrabang, HiveGround Co, Ltd., University of the Ryukyus
Sign up to receive updates directly to your inbox.
