| Rape,the largest oil crop in China,plays a crucial role in agricultural production.Moisture stress can cause delays in its development,increase the likelihood of disease,and decrease yield.The determination of moisture content is crucial for analyzing the health of the crop and guiding water irrigation,making it highly significant for the growth of the oilseed industry.The conventional method for moisture content detection is through the drying method,which is accurate but time-consuming and destructive to the test sample.Spectral detection technology offers fast,non-destructive,and accurate measurement,and has been successfully applied to the moisture detection of crop leaves.This article describes the use of a hyperspectral imaging system to obtain spectral data of rapeseed leaves,extract feature bands with high correlation to leaf moisture,and develop a portable,non-destructive,and accurate instrument for detecting rapeseed leaf moisture.The instrument uses a partial least squares regression prediction model and a reflected light path with a CMOS photoelectric sensor.The main research contents and conclusions are as follows:(1)Employ soilless cultivation techniques for rapeseed plants,establishing varying water gradients within the experiment,and acquire rapeseed leaf sample data for each gradient.Utilize a visible near-infrared hyperspectral imaging system to gather hyperspectral images of the samples.The entirety of the rapeseed leaf serves as the region of interest(ROI),with the average spectrum of all pixels in the ROI constituting the original spectrum of the sample.Determine the rapeseed leaf’s moisture content through the drying method.Perform Savitzky-Golay smoothing filtering on spectral data,and extract 11 feature wavelengths(484,517,539,558,671,684,708,720,741,885 and 965 nm)utilizing the continuous projection algorithm(SPA).(2)Evaluate device selection and structural design for blade moisture detection instruments.Investigate application scenarios and spectral characteristics of various light sources,opting for micro low-power halogen light sources as light source generation components.Examine the principles of photoelectric sensors and employ CMOS photoelectric sensors as spectral data output devices.Design a leaf spectral measurement fixture with black/white board switching to accommodate the physiological characteristics of thin,flat crop leaves.Develop a portable instrument structure for detecting moisture content in rapeseed leaves based on a reflected light path.(3)Design software and hardware for leaf moisture detection instruments based on photoelectric sensors.Hardware design encompasses power consumption estimation,power supply layout distribution,peripheral circuit construction of photoelectric sensors,high-speed analog-to-digital conversion circuit,digital logic circuit of the main control chip,and printed circuit board fabrication.Software design includes interface display,button control,photoelectric sensor spectral data collection,and more.(4)Conduct functional verification and performance testing of the leaf moisture detector.During the software and hardware debugging phase,verify the output intensity and stability of the light source,the voltage accuracy and stability of the circuit board,the accuracy and stability of analog-to-digital conversion data,and the linearity and stability of the photoelectric sensor data.Employ the least squares fitting method to calibrate the photoelectric sensor’s wavelength and ascertain the value range of the reflection spectrum data in the characteristic band.(5)Under varying moisture gradients,a total of 150 rapeseed leaves were selected as experimental samples,with spectral data at characteristic bands measured using specialized instrumentation.The true moisture content was ascertained utilizing the drying method.The training set and test set are divided by 3:1 ratio,and partial least squares regression(PLSR)prediction algorithm is used to establish detection model based on seven characteristic wavelength of 671,684,708,720,741,885,and 965 nm.The RC and RMSEC were 0.9078 and 0.1538,the RP and RMSEP were 0.8917 and0.1387.In summary,the instrument developed in this study enables non-destructive,convenient,and accurate detection of the moisture content in rapeseed leaves. |
