Establishment And Application Of UAV Imaging Hyperspectral Cotton Spectral Library

Cotton is an important strategic resource in my country,and the real-time monitoring of their growth conditions is related to the income of cotton farmers and the development of the rural economy.Using unmanned aerial vehicle(UAV)equipped with imaging hyperspectral sensors to monitor cotton and analyze its spectral data can accurately and quickly obtain nutrients,growth status and other information without damaging the crop.However,the amount of UAV imaging hyperspectral data is huge,so it is an inevitable choice that using software engineering technology to establish a hyperspectral database.It could visually manage and efficiently analyze cotton hyperspectral data.In response to this demand,this paper designed and built an UAV imaging hyperspectral cotton spectral library.It could optimize the classification accuracy and recognition accuracy in remote sensing applications in the cotton field,effectively improve the level of remote sensing quantitative analysis,provide information for the inversion of cotton growth parameters,strengthen the application of cotton hyperspectral data and the reliability of cotton remote sensing monitoring.It was expected to provide reference for UAV remote sensing and hyperspectral monitoring of crop growth.The main research contents and results were as follows:1.This article analyzed and discussed the data collection specifications,data sorting format and storage audit conditions of UAV imaging hyperspectral data.The multi-rotor UAV M600 PRO equipped with frame imaging hyperspectral sensor Rikola was used to collect the imaging hyperspectral data of cotton from the seedling stage to the boll stage.The data was collected in a typical machine-harvested cotton planting area in Moguhu Village,Shawan County,Tacheng District,Xinjiang.The data acquisition time was from May to August 2019.Then some pre-processing works such as band registration,image stitching,and radiation correction were be implemented on the UAV imaging hyperspectral data.And it recorded temperature,humidity,wind speed,collection time,cotton plant height,cotton health status,measuring instrument,spectral resolution,start and end wavelength,longitude and latitude,sea wave,collection method,collection personnel and other data as supporting attribute data.Besides,using the SR-3500 spectrometer obtained ground canopy spectrum data as verification data.It lay a data foundation for the establishment of airborne imaging hyperspectral cotton spectral library.2.In order to extract an accurate cotton spectrum curve from the preprocessed hyperspectral results images,this paper compared and analyzed the endmember extraction results of the N-FINDR algorithm and the pure pixel index algorithm(PPI)algorithm.The standard was the ground canopy spectrum curve obtained by the SR-3500 ground object spectrometer,and the index was the spectrum angle.It was analyzed from the aspects of the end-member extraction algorithm effect,the data comparison of different altitudes,and the spectral correlation.As a result,The N-FINDR algorithm’s spectral angle results at 60m,80m,and 100m altitude were 0.0658,0.0659,and 0.0677,respectively.Compared with the PPI algorithm,the N-FINDR results were closer to the data of the SR-3500 spectrometer and it had better correlation(R~2 All are above0.99),and the algorithm can better extract small sample endmembers.Secondly,there are differences in the spectrum curves which obtained by different acquisition platforms,so it is necessary to establish an UAV hyperspectral spectrum library.Thirdly,below 100m,the aerial height has a little influence on the endmember extraction,and the difference in the extraction results of the same algorithm at different aerial heights is less than 2%.3.According to the application requirements of the spectral library and the characteristics of hyperspectral data,the paper designed the system function modules,system architecture,data specifications,and database physical structure.Using relational database Oracle 11g to store and manage airborne hyperspectral images,cotton spectrum curves and supporting attribute data.Under the Visual Studio2015development platform,based on C#and IDL language,using the B/S development model,the UAV imaging hyperspectral cotton spectral library system was designed and built.It realizes the visual management and efficient data analysis of spectral data,supporting attribute data and hyperspectral images.4.The cotton hyperspectral data in the spectral library and some of the system’s spectral analysis and image analysis functions were used to analyze the changes in the spectral characteristic parameters and vegetation index of the cotton during the growth process.The results showed that the spectral characteristics of cotton and the changes in vegetation index in different growth periods were regularity.The absorption valley and the red boundary value reached their peaks in mid-July(0.0279,0.0111).The standard vegetation index and the ratio vegetation index reached their maximums in the first and mid-July(0.8417,11.6305).Enhanced vegetation index,differential vegetation index,optimized soil regulation vegetation index reached its maximum value(0.8189,0.5013,0.5012)in the middle and late July.The analysis shows that the changes of multiple spectral indicators show a parabolic trend,reaching an extreme value in July(flowering period).It indicates that the absorption of red light is the most significant during the flowering period,and photosynthesis is the strongest at that time.