Design And Evaluation Of Multispectral For Crop Nitrogen Sensor

Crop growth information monitoring is the premise of precision management of modern agriculture.Rapid and accurate access to crop nutrition information is an inevitable requirement for the sustainable development of modern agriculture in China.Nitrogen directly reflects the nutritional status of crops and is an important index for nutritional diagnosis,crop growth monitoring and quality evaluation.Economical,portable,real-time,accurate,low-cost,fast and losses nitrogen nutrition diagnosis monitor is one of the key technical means of scientific management of crop nitrogen fertilizer.In view of the problems that most crop nutrient diagnostic instruments on the market have single sensitive band,few potential spectral indexes,high price and difficult to carry out secondary development,a portable multispectral crop nitrogen diagnostic monitor is designed and developed in this study.The monitor can obtain crop growth information quickly,accurately and losslessly,and the cost is low.The results provide a reference for guiding crop field production and nitrogen management,and realize scientific and efficient fertilization.The main research results of this thesis are as follows:(1)According to the advantages and disadvantages of current crop nutrient diagnostic instrument,in order to obtain leaf nitrogen content(LNC)data quickly and at low cost,based on the existing work of the team,the design and development of crop nitrogen diagnostic monitor(MCNS)based on multispectral is carried out.STM 32,two optical devices(including visible 6-band and near-infrared 6-band),light source,ranging system,temperature and humidity sensor,key hardware equipment such as display screen,SD card,Bluetooth and power supply;The embedded control software system is designed with C language to realize the collection,display,storage and diagnosis of crop spectral information.After the MCNS design is completed,the parameters of the calibration monitor are adjusted and adjusted.The results show that the monitor can accurately obtain the spectral information of corn leaves.Compared with the same type of instruments,MCNS has the advantages of light weight,economic price,more monitoring information,and updatable upgrading.(2)In this study,MCNS,PSR+ground hyperspectral spectrometer and SPAD were used to obtain leaf spectral data at the same time.The field experiment showed that the measured values of MCNS were well correlated with those of PSR+hyperspectral spectrometer and SPAD.By optimizing the 27 spectral indexes calculated by MCNS,it shows that the best effect of predicting LNC is the three band spectral index m RER,with R~2 of 0.80.The R~2 values of the other five three band spectral indexes BNI1,PSRI,m ND705,m SR705,SIPI and a fusion band spectral index NPDI are all above 0.76.On the basis of optimizing the spectral index,in order to improve the predictability and accuracy of LNC,machine learning algorithm is introduced.Taking 27 optimized spectral indexes as input variables,training partial least squares(PLSR)and random forest(RF)models,and taking univariate regression analysis as the benchmark,the results show that compared with univariate regression,machine learning algorithm can improve the prediction accuracy R~2 of LNC from 0.80 to 0.89,and RMSE is significantly reduced.After ranking the importance of 27 optimized spectral indexes,the best combination of 7 optimized spectral indexes(PSRI,BNI1,m RER,BNI2,NDDA,m SR705 and SIPI)was found on the premise of the same prediction accuracy.In the comparison of two machine learning LNC modeling methods,PLSR model performs best,and the established model has good estimation ability in maize LNC monitoring.Based on the above results,this study designed and developed a fast and nondestructive crop nitrogen diagnosis monitor based on multispectral.The instrument is portable,highly integrated and low-cost.It solves the problems of single sensitive band,high price and no secondary development of most monitors.It can be easily applied to agricultural production and realize the nitrogen nutrition diagnosis of corn leaves,The timeliness of crop nitrogen diagnosis and monitoring based on near ground remote sensing is improved.