| Hyperspectral remote sensing technology can fast, non-destructive and accurate monitoring of crop growth, widely used in precision agriculture. Rape is an important oil crops in China, the use of hyperspectral remote sensing in the field of non destructive, rapid, real-time monitoring of rapeseed growth is very necessary. This test under different fertilizer rate, planting density level by studying the hyperspectral reflection characteristic of rape canopy and rape leaf net photosynthetic rate, leaf chlorophyll content, plant nitrogen content, rapeseed oil content related to agriculture parameters, analyzes the rape canopy spectral characteristics under different fertilizer density levels and related agriculture parameters with high spectral quantitative relationship, the monitoring model is established, and the inspection of the relevant data in different year. The main research results are as follows:(1) Different growth period, different fertilizer treatment, rape canopy spectral reflectance curve has the same change trend. Within 350-680nm, various treatments of canopy reflectance are relatively low, at 550nm in reflection peak, in 680-760nm, reflectivity has risen sharply, in 760-1350nm, and infrared regions, rape leaf reflectance in a high level; rapeseed red side show "Twin Peaks" feature, and the peaks are located in 720nm and 760nm location nearby, with the development of growth period, the "two peaks" phenomenon is more obvious.(2)Leaf net photosynthetic rate monitoring model based on all band reflectance model generally better than the single band spectrum parameters monitoring model,high fitting decision coefficient coefficient of 0.7 above, and multiple models of 0.75 or more. Including five model built based on all band reflectance validation fitting decision coefficient is greater than 0.7, the relative root mean square error RRMSE are less than 1, shows anticipation of the five models of rape leaf net photosynthetic rate has higher precision.Based on DVI (697,350) of the linear model y=44.26+-266.01x, equation fitting coefficient of maximum,RRMSE of minimum,can be considered to be the best monitoring model of rape leaf net photosynthetic rate.(3)Leaf chlorophyll content monitoring model based on all band reflectance spectrum parameters fitting optimization decision coefficient is generally higher than that of monitoring model based on single band spectrum parameters, including five monitoring model fitting decision coefficient is greater than 0.5, through F test, reached extremely significant level. Index quantitative model based on RSI (1580,147) y= 33.96e 17x, fitting decision coefficient, model validation fitting coefficient are maximum, root mean square difference is minimum, can be considered to be the best monitoring model of rape leaf chlorophyll content.(4) Based on the optimization of single band, rape plant nitrogen content monitoring model validation fitting decision coefficient is small, and the relative root mean square difference were greater than 4, shows that forecasting precision of the model is not enough; And based on all band reflectance spectrum parameters, in addition to the DVI (1534,450), the nitrogen content of building model validation fitting decision coefficient is greater than 0.6, the relative standard error less than 1, achieved the high estimation accuracy.(5) The rapeseed oil content monitoring model based on all band spectrum parameters fitting decision coefficient is small, less than 0.5, the explanation of the model fitting is not high. And based on R461 ’polynomial model, based on R911 linear, exponential and polynomial model and based on the DVF (461,911) model of polynomial fitting decision coefficient of 0.52 above, which is based on R911’linear, index and polynomial model validation of the R2 were greater than 0.6, RRMSE are less than 1, that model has a certain accuracy, can be used as a monitoring model of rapeseed oil content. |
PDF Full Text Request