Monitoring Of Grain Filling And Maturity Of Winter Wheat Based On Multi-source Remote Sensing Imagery

Wheat filling period is a key period of the formation of yield and quality.Timely and accurate monitoring of grain filling period is important for high-yield wheat cultivation.Traditional field survey methods are time consuming,Laborious and destructive.With the development of agricultural modernization,remote sensing technology is increasingly applied to agriculture.RGB digital images have advantages of lowest cost and widely application.In addition to color,water content is also a typical indicator of wheat filling and maturity.However,RGB digital images cannot express water information because it only contains visible light bands.Therefore,this study used thermal infrared data to express water information by temperature,and combined with color information of digital imagery to study wheat grain filling and maturation.This study uses RGB digital data and thermal infrared data to focus on the grain filling and maturity monitoring of wheat ears.It aims to provide theoretical reference and technical support for field management and timely harvesting during grain filling.In order to accurately monitor the grain filling and maturation of wheat ears,this study used digital image data to identify wheat ears.The wheat ears from field digital image were accurately segmented by color features based on the super-pixel segmentation.On this basis,the synchronously acquired thermal infrared image was masked to extract the wheat ear temperature data.In order to quantitatively monitor the grain filling process and maturity of wheat,the inversion model of days after anthesis and the maturity index of wheat ear at the end of grain filling were constructed based on the color and temperature characteristics.The main results are as follows:1)The color features used in this paper can effectively identify wheat ears during grain filling.Compared with the pixel-based segmentation method,the wheat ear based on super pixel block segmentation had complete morphology and edge information,and the misclassification phenomenon was also improved.The verification results of wheat with different growth condition showed that the wheat ear recognition effect was the best under the high nitrogen level,and the recognition accuracy was 94.4%.Under the nitrogen-free treatment,the growth of wheat was weak,and the ear identification was the most unsatisfactory.The recognition accuracy was only 81.9%.The identification accuracy of wheat ears in the more normal mixed samples reached 92.9%,which was 8.3%higher than that of the mixed samples with more significant differences in growth.Therefore,in the general environment,the wheat ear identification method using superpixel and color features can quickly and accurately identify and segment the wheat in the field.2)The thermal infrared image data acquired at different periods have different sensitivity to water treatment.The thermal infrared data obtained around 5 pm when the solar elevation angle is approximately 24°-28° was most sensitive to the difference between water treatments and the data quality was the best.Compared with the RGB color model,the color characteristics of the LAB color model were most sensitive to the color change during grain filling.Based on Normalized ear temperature(NRET)and Days after anthesis(DAA)can accurately simulate grain water content(GWC),the correlation coefficients R2 of the models were 0.86 and 0.91,and the relative root mean square errors NRMSE were 0.07 and 0.09,respectively.On this basis,the estimation of days after anthesis by normalized ear temperature index combined with color characteristics also achieved ideal results.RMSE and NRMSE were 2.31 days and 0.09 days respectively,which could be applied to monitoring the grain filling process of wheat.3)The yellow and red values under LAB color model could be used to characterize the maturity of wheat ears.The senescence index based on color features(SIc)could be used to judge whether wheat has entered the mature stage,and SIc?1 represents maturity.For the mature stage,the normalized ear temperature index(NRET)was used as the characteristic index to construct the Maturity index(MI)to quantify maturity.When MI was equal to 0.8,it was the best harvest period for wheat.When the MI was greater than 1,the wheat entered the ripening stage.During the period,production was easy to lose.In the early stage,only digital images were used for analysis using sectional evaluation,which could reduce the monitoring cost to the greatest extent and provide a new technical reference for the timely harvest of precision agriculture.