Research On Portable Near-infrared Spectroscopy Fast Detection Technology Based On FPGA+ARM

With the development of society and people's increasing emphasis on healthy diet,crops containing high protein are becoming more and more popular.At present,the devices used for on-site inspection of crops are large in size,difficult to carry,and relatively expensive.This paper studies the rapid detection technology of the component content of granular crops.Through the combination of near-infrared spectroscopy and embedded technology,a portable device dedicated to the rapid detection of the component content of granular crops is designed to realize the granular crops.On-site real-time detection of component content meets the market's demand for portable rapid detection devices.First of all,this article investigates the market demand in depth,analyzes the reasons why the near-infrared spectroscopy technology is effective in crop detection,and expounds the significance of designing a portable measuring instrument for the component content of granular crops in actual production and life.The main types of portable near-infrared spectrometers and their development status at home and abroad.Secondly,through the analysis of the working principle of the near-infrared spectrometer,this paper proposes the overall design scheme of the hardware of the portable near-infrared spectrometer with FPGA+ARM architecture.The design ideas and working principles of each module in the overall design plan are introduced in detail.The plan has the advantages of low cost,rich functions,and simple development.Taking into account the application range and price of the device,the corresponding CMOS device,ADC device and FPGA chip were selected,and the design of CMOS drive module,ADC drive module and FPGA control module was completed,and the timing of each module of the design was carried out.Simulation,and finally completed the development of the overall hardware.Finally,this paper uses the self-developed spectrometer equipment to collect the absorbance data of the soybean sample set,and establishes a calibration model for soybean protein,fat and moisture.The PCA method was used to complete the elimination of abnormal samples,and the abnormal values in the twenty groups of soybean samples were successfully eliminated.After comparing six different pretreatment methods,it is known from the pretreatment results that the combination of df₁+S-G+SNV pretreatment has the best effect.Comparing two different modeling methods of PLSR and PCR,the results show that the modeling effect of PLSR is better than that of PCR.The parameter results of the final modeling are imported into the ARM program to realize the integrated integration of the spectrometer.The test result of the spectrometer shows that the absolute error between the test result of the soybean component content and the test result of the standard chemical method is controlled within 3%.The causes of errors and related factors affecting accuracy are analyzed,and key factors that need to be improved are pointed out.