Font Size: a A A

Multi-platform Remote Sensing Monitoring Research Of Plant Nutrition And Stress

Posted on:2017-11-27Degree:MasterType:Thesis
Country:ChinaCandidate:X Y AiFull Text:PDF
GTID:2370330548483757Subject:Photogrammetry and Remote Sensing
Abstract/Summary:PDF Full Text Request
China,as the world's largest grain and fruit production,use remote sensing method in real time,fast and non-destructive to obtain large areas of farmland and orchard information,monitoring plant growth status,it is significant for the realization of the high yield,quality,efficiency,ecology and safety of agriculture in China.This study takes changping district,xiaotangshan of Beijing precision agriculture demonstration base as the study area.By carrying out the investigation of the nutrient and the stripe rust of wheat in the ground,the canopy spectrum and the active type of leaf physiological observation are observed in the multiple critical growth stages.Then,the nutrition and the disease investigation of the crops are carried out;A rotor UAV remote sensing monitoring system equips with Lightweight Agricultural Digital Camera(ADC-Lite)multispectral sensor is designed to carry out the test.In order to obtain the maturity,of the multi-spectral remote sensing image of peach orchard.The physiological and biochemical parameters of peach including Chlorophyll relative content(Chl),anthocyanin index(ANTH),nitrogen balance index(NBI)and nutrition and the disease investigation are monitored by remote sensing.Spectral derivatives includes vegrtation indices.Moreover,textual information,including 8 grey-level co-occurrence matrix(GLCM)indices,is extracted using four window sizes including 33?55?77?99.The optimized spectral features and texture features of different scales are combined with the physiological and biochemical parameters of the ground.The physiological parameters inversion model is constructed by partial least squares regression(PLSR).The model accuracy evaluation index is the decision factor(R2)and the root mean square error(RMSE).Then,the highest accuracy of feature combination is the optimal model.The results shows,nutrient hyperspectral monitoring of crops in near surface:by comparing and analyzing the monitoring model of the physiological and biochemical parameters of wheat,which were constructed with hyperspectral vegetation index and multi spectral vegetation index.the model of the high spectral vegetation index is the highest accuracy,the highest inversion accuracy of Chl is in the anthesis stage,R2 is 0.97,RMSE is 1.7,the highest inversion accuracy of Flav(flavonoids)is in the anthesis stage,R2 is 0.95,RMSE is 0.039,the highest inversion accuracy of NBI is in the jointiong stage,R2 is 0.89,RMSE is 2.1.Disease hyperspectral monitoring of crops in near surface:based on the optimal spectral features and based on leaf physiological observations Flav(flavonoids),Chi of different combinations on wheat flag,early filling and grain filling stage respectively has a better performance,precision of the model R2=0.90,RMSE=0.026.Compared to pure using spectral characteristics,comprehensive canopy spectra and leaf physiological observations improves model accuracy by 21%,shows that the combination of the two kinds of data can improve the illness severity estimation precision.Nutrient monitoring of orchard with UAV,the model of the spectral features the texture features of different scales of UAV multi spectral image improvs the accuracy of the single spectral-based or single texture-based of the estimation of Chi,ANTH.But for NBI,the improvement accuracy is not significant.Compared to the capability of single spectral-based models for the physiological and biochemical parameters,the models derived from single texture,texture and spectral provide the better estimates;Compared to the capability of single spectral-texture models for the physiological and biochemical parameters,the advantage of the combination of spectral and texture are not obvious.Moreover,the optimal models of Chi ANTH respectively derived from spectral-based and texture-based with 5x5 and 3x3 window size.R2 respectively is 0.897,0.414,and RMSE respectively is 1.83,0.0427.The distribution map of Chl,ANTH of the mature stage could reflect the true nutritional status of peach trees.Disease monitoring of orchard with UAV,the model of the combination of texture features and spectral features has the best effect,and the window size of the highest accuracy of the model is 77,and the forecast accuracy is 63%.Therefore,remote sensing monitoring based on different platforms is feasible.
Keywords/Search Tags:Plant, Vegetation Index, PLSR, Unmanned aerial vehicle, GLCM
PDF Full Text Request
Related items