Research On Cotton Drought Identification And Grading Based On Hyperspectral And UAV Imagery

Soil moisture is the key factor required for cotton growth and development,under the condition of water stress,the morphology and physiological indicators of cotton will be affected to varying degrees,thereby affecting the normal growth and yield of cotton,the use of remote sensing method for real-time monitoring and rapid diagnosis of cotton moisture status,to improve cotton field irrigation measures,improve cotton field water use efficiency plays an important role.In this test,Tahe No.2 cotton was used as the test material,and five irrigation water gradients were set up,which were 900 m3·hm^-2,1 800 m3·hm^-2,2 700 m3·hm^-2,3600 m3·hm^-2,and 4 500 m3·hm^-2.In five different periods:cotton bud stage,first flowering stage,full bloom stage,early bell stage and full bell stage,the hyperspectral reflectance of cotton canopy and UAV multispectral images were obtained,and the functional leaves of cotton were collected,some physiological indexes were measured,and the physiological indexes,spectral reflectance and characteristics of UAV multispectral images of cotton under different irrigation amounts in different growth periods were analyzed.Using hyperspectral and UAV multispectral parameters,partial least squares regression（PLSR）and support vector machine regression（SVM）were used to estimate and model some physiological indexes of cotton,and the drought in cotton fields was identified and graded according to the estimation model.The main conclusions are as follows:（1）Through the position relationships of the three sides of the spectrum,red valley and green peak,the spectral curve characteristics under different water treatment in different periods were analyzed,and it was found that the yellow edge position at the peak of the bell was more sensitive to water changes,the wavelength changed between 590 nm-629 nm,and the yellow edge position had a tendency to move to the long-wave direction with the increase of irrigation volume,and the yellow edge amplitude of cotton under water stress in the bud stage was very different from the normal irrigation volume（P<0.01）,while the absorption area and absorption width of the absorption valley near 500 nm.It becomes larger with the increase of irrigation volume.（2）The chlorophyll,nitrogen,phosphorus,potassium and calcium content of cotton decreases with the decrease of irrigation volume;With the advancement of cotton growth and development,the content of these physiological indexes in cotton leaves is also increasing.Analysis of variance showed that the content of these physiological indexes in cotton leaves was significantly different under different irrigation water conditions in the same growth period（P≤0.05）.（3）The spectral parameters of the optimal pretreatment were selected,and some physiological index estimation models based on PLSR and SVM were constructed in each growth period of cotton.（4）The vegetation index constructed by UAV multispectral band was combined with the physiological index of cotton leaves,and an estimation model based on SVM was established.Among them,the water content,nitrogen content,potassium content and calcium content were verified by accuracy,and the spatial distribution maps of the four indicators of each growth period were made according to the model.（5）According to the spatial distribution map of water content,nitrogen,potassium and calcium,the weight of each index was calculated,so as to construct the drought index and complete the production of the cotton drought grading map.