Research On The Application Of Silicon-based Spectroscopic Chips In The Detection Of Fruit Brix

Fruit’s soluble solid content(SSC),also known as sugar content,is an important indicator to measure the taste and quality of fruit,and also determines the time of fruit picking.In recent years,spectral detection technology has been widely used in the rapid nondestructive detection of fruit sugar content,which has the advantages of no pollution,fast detection speed and accurate results.However,traditional spectral detection equipment is difficult to enter civil life due to its high cost and large volume.In order to better meet the needs of scientific research and civil use,silicon-based spectral chips have replaced traditional detectors with obvious advantages of small size,low cost and high signal-to-noise ratio.In this paper,silicon-based spectral chip is used as the detection device,and high-precision scientific research spectrometer is used as the comparison device,taking the sugar content of Shandong Fuji,Shaanxi Fuji,Xinjiang Aksu and Crown pear as the research object,based on the multi-spectral imaging technology and visible-near-infrared spectroscopy analysis technology,combined with the stoichiometry method,the universal detection model of sugar content of various fruits was established and optimized.The main research contents of this paper are as follows:(1)The prediction of sugar content of different fruits(apple,pear)by visible-near infrared spectroscopy was studied.The diffuse reflectance acquisition platform was built to collect the diffuse reflectance spectra of different fruits and convert them into absorbance.In order to improve the prediction accuracy and stability of the model,combined with partial least squares(PLS),multivariate scattering correction was selected as the best spectral pretreatment method,and regression model of sugar content of many kinds of fruits was established based on PLS,coefficient of determination of the model_pR~2 is 0.88,and the root mean square error of prediction RMESP is 0.545.Random forest(RF)was used to predict the sugar content of various fruits.Compared with PLS regression model,the results show that the performance of RF regression model is better than that of PLS regression model,_pR~2 was 0.90,RMSEP was 0.487.The overall effect of the model is improved.(2)The feasibility of silicon based spectrum chip in the determination of sugar content in fruits of the same type and many kinds was explored.Based on multispectral imaging analysis technology,multispectral image of fruit sugar content after image processing,after extracting spectral data,PLS and RF regression models were constructed and compared.By comparison,in the prediction of sugar content of various fruits,the prediction set RF regression model_pR~2 was 0.87,and RMESP was 0.473.The results showed that the application of silicon spectrum chip in fruit sugar content detection is feasible.In conclusion,the visible-near infrared detection models of fruit sugar content by spectrometers were compared,based on multispectral imaging analysis technology,silicon based spectrum chip can be used to achieve rapid and nondestructive detection of fruit sugar content.It has theoretical and reference significance for nondestructive testing of fruit picking and grading.At the same time,it has practical significance and reference value to realize the civil application of silicon based spectrum chip in fruit sugar content detection as soon as possible.