Research On Cotton Growth Parameter Monitoring Model Based On UAV Remote Sensing Images

As an important Cash crop in China,cotton has the characteristics of strong a daptability and wide planting range.The growth parameters of cotton are important index to indicate its health status.Rapid and non-destructive monitoring of the gro wth parameters of cotton is the prerequisite for stable and increased production of cotton.The traditional manual sampling method is too time-consuming and costly to meet the needs of accurate management.The use of UAV remote sensing technolo gy to monitor the growth of cotton at key fertility stages provides technical support for the precise management of cotton in the field.A remote sensing experiment for cotton growth monitoring was conducted in Y angliuxue Town,Binzhou City,Shandong Province,from May to August 2021,usin g different varieties of cotton breeding materials as research objects.The UAVs wer e equipped with visible and multispectral cameras to acquire remote sensing image data during four key fertility stages of cotton:seedling,bud,first flower and boll,and to collect agronomic parameters such as relative chlorophyll content,leaf area i ndex and plant height simultaneously on the ground.The remote sensing data from ground(canopy analyzer,chlorophyll meter)and UAV(DJI Genie 4 RTK visible li ght camera,DJI M210 with YUSENSE multispectral camera)were combined to res pectively construct the SPAD and LAI estimation models of cotton at critical fertilit y stages based on the univariate linear regression model,stepwise regression model and artificial neural network model.The optimal SPAD and LAI estimation models were selected through comparative analysis of the above models,using visible light remote sensing data to estimate plant height at key fertility stages of cotton.(1)Estimation of cotton SPAD based on UAV visible light and multispectral i magery-The cotton SPAD was estimated based on visible vegetation index,multis pectral vegetation index and visible+multispectral vegetation index,respectively,an d analyzed using a univariate linear regression model,stepwise regression model an d artificial neural network model.The results showed that the artificial neural netw ork model constructed using the visible vegetation index had the best prediction acc uracy,with the modeling set coefficient of determination R~2 of 0.849 and root mea n square error of 1.1172,and the validation set coefficient of determination R~2 of0.919 and root mean square error of 0.8868.(2)Estimation of cotton LAI based on UAV visible light imagery and multispe ctral imagery-Estimation of cotton LAI based on visible vegetation index,multisp ectral vegetation index and visible+multispectral vegetation index respectively,corr elation analysis using the extracted visible vegetation index and multispectral vegetat ion index with the ground measured leaf area index,using the preferred vegetation index to construct univariate linear regression model,stepwise regression model and artificial neural network model.The results showed that the artificial neural network model based on the visible vegetation index had the best prediction accuracy,with the modeling set coefficient of determination R~2 of 0.959 and root mean square er ror of 0.2818,and the validation set coefficient of determination R~2 of 0.913 and r oot mean square error of 0.3767.(3)Estimation of cotton plant height based on UAV visible light images-Cott on plant height in the experimental area was extracted using the digital surface mo del DSM,and cotton plant height was estimated separately according to single and multiple fertility stages.The results showed that the highest coefficient of determina tion R~2 for single fertility stage was 0.686,with a low prediction accuracy,and the accuracy of cotton plant height estimation for multiple fertility stages is higher,th e coefficient of determination of the model was 0.958.In summary,this paper uses UAVs equipped with visible and multispectral cam eras to acquire remote sensing images and analyse the chlorophyll relative content(S PAD),leaf area index(LAI)and plant height measured in the field of cotton breedi ng materials.Through the quantitative prediction model constructed,the objective an d rapid acquisition of chlorophyll relative content,leaf area index and plant height is achieved,and this paper can provide reference for the accurate monitoring of cot ton growth parameters using remote sensing technology.