| The passive crop growth monitoring equipment can only be used at the cloudless and sunny day from 11:00-14:00.In order to overcome these shortcomings,both domestic and foreign research institutions have developed some active crop growth monitoring equipment,which are suitable for all weather conditions.The National Engineering and Technology Center for Information Agriculture has developed the active crop growth monitoring instrument CGMD402.However,the CGMD402 has not been systematically tested and evaluated in field experiments,moreover,the crop growth models of the CGMD402 are required to satisfy the demands of precision crop growth management.Therefore,this study was based on the CGMD402,mainly focus on the equipment performance evaluation.Various rice and wheat experiments are designed to testify the consistence of using CGMD402 at different time,different orientations and different heights.As a comparison,The ASD and RapidSCAN were tested at the same time during all crop growth stages.Finally,a crop growth spectrum monitoring model was constructed with LAI(Leaf Area index),LNC(Leaf Nitrogen Content),LDW(Leaf Dry Weight)and SPAD(Relative chlorophyll content).The following work was mainly completed:1.The performance experiments were implemented at both Huaian and Rugao in Jiangsu province,CGMD402 was tested in Huaian and Rugao for different nitrogen levels of rice canopy spectra,the results shown that(1)The variation coefficient of experiments about different orientation,height and time were 1.6%,3.6%,4.7%,respectively,signal-to-noise ratio were 77.72,60.38,51.4,respectively.The measuring orientation and time will not affect the performance of CGMD402,and CGMD402 is relatively stable at the height of 20-60cm.Therefore,CGMD402 should be kept at 20-60cm upright from the crop canopy height.(2)The change characteristics of vegetation indexes(NDRE and RVI)measured by CGMD402 with growth stages as follow:The NDRE and RVI showed the trend of "lower-higher-lower" during the growth period.From tillering to booting,the value of them gradually increased and reached the peak;From booting stage to heading stage,they both decreased slightly.The difference of vegetation indexes between the two rice varieties decreased gradually with the growth period.By contrast,the difference of vegetation indexes between different nitrogen levels increased with the growth period.Under different planting densities,the difference of vegetation indexes increased and then decreased.With different nitrogen treatments,planting densities and rice varieties,the vegetation indexes measured by CGMD402 are correspondent with the growth and change rules of rice.(3)NDRE output of the CGMD402 had a linear fit coefficient of determination(R2)of 0.8167,and a root mean square error(RMSE)of 0.218 compared with the ASD;to compare with RapidSCAN,the R2 and RMSE had changed to 0.6868 and 0.288,respectively.RVI output of the CGMD402 had a linear fit coefficient of determination(R2)of 0.8428,and a root mean square error(RMSE)of 0.201 compared with the ASD;to compare with RapidSCAN,the R2 and RMSE had changed to 0.702 and 0.284,respectively.2.The regression coefficients between NDRE from rice canopy reflectance measured by CGMD402 and four growth parameters(LNC、LAI、LDW、SPAD)could be calculated by the data of two rice field experiments in Huaian(R2)of 0.614,0.6627,0.6625,0.7714.The regression coefficients between RVI and four growth parameter(LNC、LAI、LDW、SPAD)are also produced R2 of 0.6159,0.6568,0.6567,0.7651.Then this study used the independent experiments from 2017,2018 years to test those models,and calculated R2,RMSE and RE between measured values and predicted values.The NDRE results that R2 are 0.7132,0.6958,0.7651,0.7227,RMSE are 0.274,0.271,3.15,0.271,RE are 0.15,0.11,0.184,0.15.The RVI results that R2are 0.6782,0.7839,0.7265,0.7292,RMSE are 3.05,0.81,0.55,4.3,RE are 0.148,0.11,0.12,0.15.Results showed that the rice growth monitoring model based on CGMD402 is universal and accurate,the model can be more accurate to crop growth index(LNC,LAI,LDW,SPAD)for inversion,the highest accuracy in rice monitoring is LAI,followed by SPAD,LNC,the last is LDW.3.The performance experiments were implemented at both Rugao in Jiangsu province,CGMD402 was tested in Rugao for different nitrogen levels of wheat canopy spectra,the results shown that(1)The variation coefficient of experiments about different orientation,height and time were 2.2%,4.741%,8.02%,respectively,signal-to-noise ratio were 57.14,55.06,54.64,respectively.The measuring orientation and time will not affect the performance of CGMD402,and CGMD402 is relatively stable at the height of 20-60cm.Therefore,CGMD402 should be kept at 20-60cm upright from the crop canopy height.(2)The change characteristics of vegetation indexes(NDRE and RVI)measured by CGMD402 with growth stages as follow:The NDRE and RVI showed the trend of "lower-higher-lower" during the growth period.From tillering to booting,the value of them gradually increased and reached the peak;From booting stage to heading stage,they both decreased slightly.The difference of vegetation indexes between the two wheat varieties decreased gradually with the growth period.By contrast,the difference of vegetation indexes between different nitrogen levels increased with the growth period.Under different planting densities,the difference of vegetation indexes increased and then decreased.With different nitrogen treatments,planting densities and wheat varieties,the vegetation indexes measured by CGMD402 are correspondent with the growth and change rules of wheat.(3)NDRE output of the CGMD402 had a linear fit coefficient of determination(R2)of 0.7173,and a root mean square error(RMSE)of 0.250 compared with the ASD;to compare with RapidSCAN,the R2 and RMSE had changed to 0.8003 and 0.203,respectively.RVI output of the CGMD402 had a linear fit coefficient of determination(R2)of 0.7396,and a root mean square error(RMSE)of 0.215 compared with the ASD;to compare with RapidSCAN,the R2 and RMSE had changed to 0.7857 and 0.219,respectively.4.The regression coefficients between NDRE from wheat canopy reflectance measured by CGMD402 and four growth parameters(LDW,LAI,LNC,SPAD)could be calculated by the data of two wheat field experiments in Rugao(R2)of 0.7103,0.7671,0.6625,0.7714.The regression coefficients between RVI and four growth parameter(LDW,LAI,LNC,SPAD)are also produced R2 of 0.7165,0.7949,0.7385,0.7651.Then this study used the independent experiment from 2018 years to test those models,and calculated R2,RMSE and RE between measured values and predicted values.The NDRE results that R2 are 0.6996,0.8728,0.7111,0.7937,RMSE are 4.87,0.439,0.56,2.14,RE are 0.16,0.09,0.15,0.12.The RVI results that R2 are 0.6914,0.8848,0.7256,0.7955,RMSE are 4.22,0.318,0.55,2.32,RE are 0.16,0.08,0.148,0.12.Results showed that the wheat growth monitoring model based on CGMD402 is universal and accurate,the model can be more accurate to crop growth index(LNC,LAI,LDW,SPAD)for inversion,the highest accuracy in rice monitoring is LAI,followed by SPAD,LNC,the last is LDW. |
PDF Full Text Request