Rice Maturation And Harvest Time Monitoring By Hyperspectral Reflectance

Rice maturation and harvest time are the major factors in the milling quality.Accurate determination of the optimal harvesting date by determining grain moisture and plant pigments should be very valuable for rice growers because harvesting at incorrect time could cause substantial loss. In recent decades, some hyperspectral approaches have been developed in accurately estimating plant pigments, nitrogen and plant water status at leaf and canopy scale, which suggests that hyperspectral remote sensing can be potentially applied in rice harvest monitoring. The objective of this work is to investigate the change pattern in the hyperspectral reflectance of rice plants during ripening and the relationships among the plant spectral features, pigments, water contents and rice milling quality.A field experiment was carried out to investigate the relationships between the hyperspectral reflectance (350-2 500 nm) of the flag leaves and the panicles, and their pigments (chlorophyll a, b and carotenoid) contents, as well as the water content and rice milling quality, in rice plants of two cultivars under three nitrogen levels from milk to harvesting. Some existing hyperspectral vegetation indices were analyzed in conelation with plant water content, chlorophyll, carotenoids in the flag leaves and panicles, and some statistical models were established to explore the feasibility of rice maturation and harvest time monitoring.The reflectance of the flag leaf and panicle over the visible and near infrared region tended to increase with the harvest time. The red edge of the first derivatives based on spectral position in the flag leaves, shifted toward shorter wavelength along with the harvest time. At ripening, the panicle water content tended to be close to 24% across the three N level reatments and the two cultivars. The rice milled rate and head rate increased with the harvest time, until at complete mature, and after then decreased. The results in this study indicated that 24%²5% could be considered as the optimal harvested grain water content for rice milling quality. Hyperspectral vegetation indices RDI₁₄₅₀, WI/NDVI were best related to panicle chlorophyll and water content, with R²=0.8098, R²=0.8798 (n=30) , respectively, which showed that these indices could be used to predict rice panicle water content, and considered as a sensitive indicator of rice maturation and harvest time. The mSR₇₀₅ and mND₇₀₅ indices were best related to the chlorophyll contents under both the low2chlorophyll and the mid-high chlorophyll level, with R2=0.9319 (n=48) for the leaves and R =0.9488 (n=48) for the panicles, respectively. PRI was negatively significantly correlated with the Car/Chl ratio in both the flag leaves and the panicles (R2=0.7745. n=48), indicating that PRI could be used to predict Car/Chl across different vegetation structure. /?76o^5oo was also significantly (p=0.05) correlated with the leaf Car/Chl, and the panicle carotenoid content. The results suggested that RDIuso, WI/NDVI, mSR705, mND705 and PRI could be potentially used for rice maturation and harvest time monitoring.