Comparison Between Rice Plant Traits and Color Indices Calculated from UAV Remote Sensing Images

Remote sensing technology for monitoring plant trains has a huge potential to accelerate breeding process. In this paper, we have studied on remote sensing of using an unmanned aerial vehicle (UAV) system for plant traits phenotyping in rice. The images of rice canopy were taken by a RGB camera from the UAV at three growing stages; Vegetative (VG), Flowering (FW) and Grain filling (GF). Typical color indices (r, g, b, INT, VIG, L*, a*, b*, H) were calculated by image processing. Single regression analysis was conducted between rice plant traits (leaf area index (LAI), grain yield, above ground biomass, plant height, panicle length, grain filling rate, tiller number) and color indices. The index a* at FW and GF had close liner relationships with LAI (the coefficient of determination R2 > 0.70) and grain yield (R2 > 0.50). Moreover, a* and g at FW and GF showed high R2 with plant height and grain filling rate (R2 > 0.50). The R2 between grain yield and color indices increased above 0.5 for about 40% of models at three growing stages by multiple regression analysis. In particular, the models of H and INT and of H and L* at VG were closely related (R2 > 0.70). Our findings show the analysis of color images taken by UAV remote sensing is useful to assessing four rice traits; LAI, grain yield, plant height and grain filling rate at early stage, and especially more available for grain yield estimation