| Grains occupy a great historical position in the survival and development of mankind.They are important substances to solve the problem of food and clothing.The correct estimation of grain yield between paddy fields is very important in precision agriculture because the yield data is mainly affected by grain moisture content.The online real-time detection of grain moisture content in the field on combine harvester is an effective way to solve the problem.At the same time,the real time understanding of the water situation of the harvested grain could control the feeding amount of combine harvester,which could reduce the mechanical working load and improve the harvest efficiency of the grain,ensure the quality of harvested grain and reduce the loss.Realizing on-line and real-time detection of grain moisture content on combine harvester is of great significance in improving agricultural economic efficiency in China.How to effectively harvest moisture content in combine harvester and detect moisture content is the main difficulty of this research.The principle prototype of vehicular grain moisture detection system using near infared spectroscopy was developed and tested in this paper based on Bilang 4LZ-2.8 type whole feeding combine harvester in the laboratory,which included the research of 1).the static detection of grain moisture content and the influence of grain thickness on the detection precision in the laboratory,2).the design of principle prototype of vehicular real-time detection system for grain moisture and 3).the dynamic detection model of grain moisture content.The main research contents and conclusions are as follows:(1)The influence of grain thickness on the accuracy of moisture near-infrared(NIR)detection was studied and the minimum detection thickness is determined.Firstly,taking the Liangyou 688 paddy as the research object,the Zeiss Corona extreme NIR spectrometer was used to detect and analyze the relationship of moisture content and detection thickness.Based on the moisture condition that might exist when paddy was harvested,the moisture gradients of the purchased grain was set.The change range of the moisture content of the sample was controlled to be about 14 to 25%,and 6 different thickness gradients(the thickness were 2 mm,5 mm,10 mm,15 mm,20 mm,and 50 mm,respectively in the container with a diameter of 100 mm,and 100 samples per thickness gradient)were detected by NIR spectroscopy measurement in different moisture content samples.The results of PLSR modeling showed that the moisture content of rice samples under various thicknesses was satisfactory,whose prediction correlation coefficient,rp,were 0.9838,0.9849,0.9845,0.9855,0.9896 and 0.9896,respectively,root mean square error of prediction,RMSEP,were 0.5183%,0.4524%,0.4277%,0.4186%and 0.3836%,respectively.Then,based on the above research data,mutual prediction analysis between different thicknesses was carried out to determine the minimum detection thickness.Spectral data with a thickness of 2 mm was used as a modeling training set for PLSR,and the remaining 5 thickness spectral data were used as modeling prediction sets to analyze the prediction results.The mutual PLSR modeling analysis results showed that the PLSR model with 2 mm thickness(single layer densely covered)of the sample as a training set was satisfactory for the prediction of the moisture content of the remaining samples,whose rp were 0.9754,0.9779,0.9748,0.9799 and 0.9782,respectively,RMSEP were 0.6502%,0.6312%,0.6386%,0.5738%and 0.5777%,respectively,RPD were 4.0470,4.1709,4.1227,4.5859 and 4.4414,respectively.That was,when the grain thickness was larger than 2 mm,the moisture content of the grain could be measured by NIR spectroscopy.This thickness could be used as the minimum thickness for real-time detection of grain moisture by NIR spectroscopy.(2)A principle prototype of vehicular grain quality real-time detection system was designed and developed.Based on the domestically produced Bilang 4LZ-2.8 full-feeding combine harvester,A principle prototype of vehicular grain quality real-time detection system that could be installed at the exit of the combine harvester grain bin,as an unloading device,was developed.The inclined self-cleaning grain detection channel was produced from the NIR spectroscopy diffuse reflectance detection method,and the inclination angle of the detection channel was determined through the grain force analysis.The spectrum acquisition trigger unit based on the cantilever beam sensor and the impact plate was developed.The overall structure design and trial production of the principle prototype were completed.(3)The NIR spectroscopy real-time detection of grain moisture was studied.Two independent dynamic tests for grain moisture content were performed on the principle prototype of vehicular grain quality real-time detection system.First,toke the paddy variety Liangyou 688 as the research object and set the sample moisture content gradients(a total of 100 samples,divided into 33 moisture content gradients,and every 3 samples was a moisture content gradient,49 to 52#these four samples were one of these moisture content gradient),The dynamic NIR spectroscopy scan of the sample was completed on the principle prototype.PLSR modelling analysis was performed based on the physicochemical reference value of the moisture content of the samples and the prediction effect of the model was good,whose rp = 0.9783,RMSEP = 0.8278%,RPD = 4.3509.Then,this model was imported into the spectrometer operating software InProcess for verification testing.Set the sample moisture content gradients(a total of 97 samples,divided into 32 moisture content gradients,and every 3 samples was a moisture content gradient,46 to 49#these four samples were one of these moisture content gradient),The real-time online NIR spectroscopy measurement of the moisture content of the sample was completed,and the error analysis was performed with the physicochemical reference value.The results showed that the root mean square error of the predicted value was RMSEP = 0.5663%and the average relative error was 2.3%,which met the accuracy requirement of on-line measurement of grain moisture.The reliability of the detection system and the modeling method,and the validity of the model were proved.It can be applied to the real-time online NIR spectroscopy detection of paddy moisture content in Liangyou 688 variety. |
PDF Full Text Request