Hyperspectral Estimation Of Main Growth Indexes And Nitrogen And Phosphorus Content Of Rice Panicle In Different Rice Varieties

With the development and popularization of remote sensing technology, ground hyperspectral technology plays a more and more important role in crop non-destructive monitoring field. This paper studied the growth and development characteristics of different rice varieties in different growth stages. And hyperspectral models for nitrogen and phosphorus content of rice panicle and growth indexes (such as SPAD, shoot biomass and LAI) of rice panicle were built through the methods such as artificial quantitative experiment, spectral analysis, model estimation and so on. And the models were validated through samples data from the validation set. The main results were summarized as follows:1. The change trends of SPAD indifferent rice varieties at different growth periods were almost the same. The SPAD slightly decreased from jointing stage to booting stage, and slightly increased from booting stage to heading and flowering stage.Then SPAD tended to be stable and gradually declined from mid filling stage. The general change trend of SPAD between different rice varieties was:Yongyou2640>Lianjing7>Huaidao5>Huaidao9>Yangliangyou6>Yundan3. The shoot biomasses of different rice varieties in different growth period were different. The shoot biomasses of Yangliangyou6 and Yundan3 were higher and quite different from other rice varieties from booting stage. The biomass of Lianjing7 remained in the lowest level. The change tendences of LAI of different rice varieties at different growth periodswere also almost the same. They reached to the maximum in booting stage to heading/flowering stage, which were consistent with the previous studies. The general changing trends of LAI of different rice varieties in different growth periods were:Yangliangyou6> Yundan3> Huaidao9> Huaidao5> Yongyou2640> Lianjing7.2. The leaf nitrogen content of different rice varieties generally showed downward trends from jointing stage to mature stage, but showed a slight increase in heading and flowering stage. The change rule wasn’t obvious among the varieties. The leaf phosphorus content of different rice varieties increased and then decreasd from jointing stage to mature stage, and reached to the maximum in booting stage. The change rule also wasn’t obvious among the varieties. The differences of the leaf angles of the first, second and third leaf from the top in jointing stage were small between different varieties, but obviously increased from booting stage. The leaf angles of Yongyou2640 and Lianjing7 in different leaf positions were larger than other varieties from booting stage to mature stage.3. The spectral reflectance of different rice varieties in different growth periods were different. The imparities of the spectral reflectance of different varieties in the same growth period were large in 750-1300nm band, and the specific imparities in different growth periods were different. The changes of the spectral reflectance from band 900nm to 1200nm of different rice varieties in jointing stage tended to be stable, and the changes were:Yangliangyou6> Yongyou2640> Yundan3> Huaidao5> Lianjing7> Huaidao9. The imparities of the spectral reflectance from band 750nm to 950nm of different rice varieties in booting stage also tended to be stable, which reflected as: Yangliangyou6> Huaidao9> Yongyou2640> Huaidao5> Lianjing7> Yundan3. The reflectance imparities of different varieties from band 750nm to 930nm in heading/flowering stage was same to which in booting stage, and the reflectance values were:Huaidao9> Yongyou2640> Huaidao5> Lianjing7> Yangliangyou6> Yundan3. The spectral reflectance of Huaidao9, Yangliangyou6, Yundan3 from band 750nm to 909nm in early filling stage were quite different from other varieties and the differences between them with other varieties were small. The spectral reflectances were: Huaidao9> Yangliangyou6> Yundan3> others. The reflectance imparities between rice varieties from band 790nm to 990nm and 1030nm to 1120nm in mid filling stage tended to be stable and the imparities show as:Lianjing7> Yundan3> Huaidao9> Yangliangyou6> Huaidao5. The reflectance of Huaidao5 from band 350nm to 750nm in mature stage was obviously lower than other varieties, and the differences between Yongyou2640, Lianjing7 and Yangliangyou6 in this band were small. The imparities between all varieties from band 750nm to 1300nm were large and unstable.4. The SPAD and its corresponding canopy spectral reflectance of rice showed asignificant negative correlationin the 604-701nm band, and the extreme value appearedin band 654nm. The correlation coefficients between SPAD and spectral characteristic parameters, such as Dy, Xg, SDy, (Rg-Rr)/(Rg+Rr) and (SDr-SDy)/(SDr+SDy)were highly significant. Building models based on these and checking, we will find that the effect of parametric combination modelwasbetter. Among that, the effect of the linear model y=724.632-1.244x1+6639.138x2based on λg and Db was the best.5. Overground dry matter weight and canopy spectral reflectance of rice presentedsignificant negative correlation from band 350nm to 686nm, and presented highly significant negative correlation from band 350nm to 500nm. The correlation coefficients between dry matter weight and spectral characteristic parameters, such as(SDr-SDy)/(SDr+SDy), λb and Dywere highly significant. Building models based on these and checking, we will find that the effect of parametric combination modelwas good, among which the effect of the linear model y=-144184.249-66.334x1+199.091x2+59.676x3+103.987x4 based on parameter λb-λg-RVI-λo was the best.6. LAI and canopy spectrum presented highly significant correlation from band 421nm to 696nm, and the correlation coefficients between LAI and spectral characteristic parameters, such as λr, Rg/Rr, (Rg-Rr)/(Rg+Rr), RVI, NDVI, RDVI and SAVI were also highly significant. Building models based on these and checking, we will find that the effect of parametric combination model was better than single parameter, among which the effect of the linear model y=-97.908+0.223x1+8790.213x2+0.135x3 based on RVI-Dy-λr was the best.7. Total nitrogen content of rice panicleand canopy spectral reflectance presented highly significant negative correlation in the near infrared band 760nm to 1300nm. The correlations between total nitrogen content of rice panicle and spectral characteristic parameters, such as λb, SDr, SDr/SDb and DVI were good. Building models based on these and checking, it found that the quadratic function model y=-1.597-34.688x2+20.486x based on DVI was the best model estimating total nitrogen content of rice panicle. Total phosphorus content of rice panicle and canopy spectral reflectance present significant positive correlation from band 614nm to 690nm. The correlation coefficients between total phosphorus content of rice panicle and spectral characteristic parameters, such as λg, SDy, Rg/Rr, (Rg-Rr)/(Rg+Rr) and (SDr-SDy)/(SDr+SDy) were highly significant, and the best model was found. It was respectively quadratic function model y=0.975+0.01x2-0.254x based on Rg/Rr.