| Precision agriculture is the basic way to realize the aims of low consumption, highefficiency, good quality, and friendly environment. Real-time information of growth state offarmland crops can be quickly obtained by remote sensing, which can provide importanttechnical support for the implementation of precision agriculture. Hyperspectral remotesensing can detect the successive spectral curve of ground objects with hyperspectralresolution in some spectral region and form the unique multi-dimensional spectrum space.Remote sensing application focuses on spreading spatial information on spectrum dimension,thereby obtaining more spectral information of object. The application of hyperspectralremote sensing can strengthen the monitoring ability of physiological ecological parametersand improve the monitoring accuracy of crop growth. Based on field experiments andregional sampling of winter wheat, this study analyzed the hyperspectral characteristics ofwheat canopy by comprehensive application of hyperspectral remote sensing, growth analysis,physiological ecology test, mathematical statistics, and wavelet transform for different sowingdensities, water treatments, species, and water and nitrogen treaments. Then the hyperspectralestimation models of chlorophyll content, percentage vegetation cover, and leaf area index(LAI) were established on the basis of vegetation index, spectral characteristic parameters,wavelet energy coefficient, and wavelet fractal at different growth stages. Finally, theestimation model of vegetation index was used to the ETM+satellite images of remotesensing. Then the actual application probability was tested, which provided the theory basisand key technology for dynamic monitoring of growth change and precise management. Themain results are as follows:(1) Changes of chlorophyll content and LAI in different species and water and nitrogenlevels were studied as well as variation of percentage vegetation cover under different sowingdensities and water levels. Additionally, the response of hyperspectral characteristics of wheat canopy was analysed. The result showed that the spectral reflectance of visible spectrumreduced at first and then increased from wheat green-turning to maturation stage. However,the spectral reflectance of near infrared spectrum demonstrated an opposite tendency. Itdecreased under the visible spectrum and increased under near infrared spectrum as thepercentage vegetation cover, leaf area index and chlorophyll content became larger. Thespectral reflectance decreased with the increase of water shortage in the visible spectrum,while on the contrary it increased in near infrared spectrum. Following the increased nitrogenlevels, there is a declinen of spectral reflectance in visible spectrum and an increase in nearinfrared spectrum in different moisture conditions. This conclusion provides the theory basisfor monitoring the growth state of winter wheat through the canopy spectral information.(2) The correlationships between canopy original spectrum, derivative spectrum and leafSPAD value, percentage vegetation cover, LAI were analuzed respectively. The originalspectral reflectance of green-turning, jointing, heading, and pustulation stages was negativelycorrelated with leaf SPAD value, percentage vegetation cover, and LAI in visible spectrum. Inthe "red stage", the negative correlation turns to a positive one. The correlation coefficients ofderivative spectrum and leaf SPAD value, percentage vegetation cover, LAI were higher thanthose of original spectrum and SPAD value in some wavebands in green-turning period,jointing, heading, and pustulation stages. There were low correlationship between originalspectrum, derivative spectrum and leaf SPAD value, percentage vegetation cover and LAI atmaturation stage. However, when generating the estimation models for leaf SPAD value,percentage vegetation cover and leaf area index, the spectral data in the maturation stagecould not be used.(3) The data including canopy spectrum of winter wheat, leaf SPAD value, percentagevegetation cover and LAI in2010, were used to establish the estimation models based onNDVI, RVI, original spectral characteristic parameters (green peak reflectance, green peakposition, green peak area, red ebb reflectance, red ebb position, red ebb area, the ratio andnormalization value of green peak reflectance and red ebb reflectance, green peak skewness,green peak kurtosis, red ebb skewness, red ebb kurtosis, the ratio and normalization value ofgreen peak skewness and red ebb skewness, the ratio and normalization value of green peakkurtosis and red ebb kurtosis), red edge parameters (red edge position, red edge amplitude,red edge area, red edge kurtosis, and red edge skewness), wavelet energy coefficient, leafSPAD of wavelet fractal dimension, percentage vegetation cover and LAI. The precision ofestablished models was tested by using the data of canopy spectrum, leaf SPAD, percentagevegetation cover and LAI in2011. The results showed that the estimation precision of leafSPAD retrieved by NDVI was greater than that retrived by RVI. The inversion effect of NDVI was better than that of RVI in the green-turning and jointing stages, when retrieving thepercentage vegetation cover of winter wheat. However, better inversion was obtained by RVIin heading and postulation stages. Therefore, RVI should have the top priority when retrievingthe LAI, but red ebb area should have the priority when retrieving the leaf SPAD inverted bythe original spectrum characteristics in the jointing stage. Red ebb skewness should be takeninto account first in reviving and heading, and postulation stages. Percentage vegetation coverinversion based on parameters of original spectrum characteristic should give priority to thegreen peak kurtosis. For the inversion of LAI, the ratio of green peak reflectance and red ebbreflectance should be the prior considerations in the jointing stage, but green peak skewnessshould be considered first in the green-turning, heading, and postulation stages. Wheninverting the leaf SPAD based on the red edge parameters, the red edge kurtosis should betaken into account preferentially. For the percentage vegetation cover inversion based on thered edge parameters, red edge kurtosis should be a prime consideration in the jointing stage.Red edge skewness should be preferentially considered at other growth stages. Red edgekurtosis should be considered first in the inversion of LAI. The inversion of leaf SPAD basedon the wavelet energy coefficients should give first consideration to low frequencyinformation. However, high frequency information should be considered first for thepercentage vegetation cover inversion. The inversion of LAI based on wavelet energycoefficients should consider low frequency information first. The leaf SPAD, percentagevegetation cover, and LAI of winter wheat could be inverted effectively by the low frequencyinformation of wavelet fractal dimension in green-turning, jointing, heading, and postulationstages.(4) The application of estimation model was evaluated based on leaf SPAD, percentagevegetation cover, and LAI of winter wheat in remote sensing data of ETM+. The resultsuggested that there was a significant correlationship between the predicted and measuredvalues when NDVI or RVI based on the constructed canopy spectrum was used to decode theremote sensing information of ETM+.
|
PDF Full Text Request