Development And Application Of A Handheld Spectrometer Based On Micro-spectrometer

In recent years,with the gradual improvement of living standards,people not only consider the appearance but also the internal quality of fruit when buying them.Therefore,a non-destructive and efficient detection method meets the needs of the fruit industry and consumers.Compared with the traditional detection methods,visible-near infrared spectroscopy is widely used for the internal quality detection of agricultural products because of its advantages of non-destructive,rapid and pollution-free.Based on visible and near infrared spectroscopy technique,a hand-held detector(ZSTU-Vis/NIR 2021)was developed with a micro spectrometer as the core device,and the effectiveness of the device on apple sugar content detection was verified.This research mainly includes the following contents:(1)A handheld spectrometer based on the C11708MA micro-spectrometer was developed.Its hardware system includes a sensor module,a processor module,a signal conditioning module,a light source module,a communication module,and a power supply module.L1005A mini halogen lamp is used as the light source,C11708MA micro-spectrometer is used as the detector,and STM32 single-chip microcomputer is used to control the device to realize the rapid collection of sample spectral information.The stability of the device was evaluated by the reference ball made of polytetrafluoroethylene.The spectra were collected several times in two power supply modes,and the change of spectral shape and amplitude was within the expected range.(2)An APP was developed based on the Android smartphone to interact with the handheld spectrometer.And a MySQL database was built through the Alibaba Cloud server to achieve data storage.The handheld detector is controlled by application software to collect the spectrum of fruit and predict the sugar content.Android smartphone has a special application interface for communication,receiving,drawing,and analyzing the spectrum.Besides,the cloud database can upload results and download model data.(3)The application verification of the handheld detector was carried out,and the prediction model of sugar content for"Fuji"Apple was established.The model was optimized by wavelength screening and pretreatment method selection.In this research,the standard normal variate was used to preprocess the characteristic spectral bands and the partial least squares regression(PLSR)model was established to analyze the sample sets.The coefficient of determination of prediction(R_p~2)and the root mean square error of prediction(RMSEP)of Set-I were 0.88 and 0.48 respectively,and the R_p~2 and the RMSEP of Set-II were 0.74 and 0.59 respectively.Based on the PLSR model of Set-I,new samples were added to the existing model to expand the application scope of the model and update the model.The R_p~2 and the RMSEP of the updated model were 0.70and 0.67,respectively.The handheld spectrometer and two commercial devices were used to test two sample sets,and the results showed that the prediction performance of the handheld spectrometer was better than the other two devices.The overall results showed that the device developed in this study is effective for the detection of apple sugar content in actual situation,and also reflects the great potential of the device for the rapid detection of fruit internal quality.