Research And Development Of Nondestructive Detection Method And Portable Device For Apple Maturity Based On Visible And Near Infrared Spectroscopy

Apple industry is one of the most important fruit industries in China.In recent years,the fresh fruit market has higher and higher requirements for the quality of apples,and consumers pay more and more attention to the internal quality of apples.Maturity is an important index closely related to the change of apple quality during ripening,which affects its yield and final quality.The existing maturity detection methods are mainly harmful detection with high labor intensity,while near-infrared spectroscopy can realize the rapid and non-destructive detection of apple maturity,which is of great significance to improve the fruit quality during apple harvest and storage.it is helpful to promote the high-quality development of the apple industry.Therefore,this paper used near-infrared spectroscopy as a tool to study the non-destructive testing method of apple maturity and developed a set of portable testing equipment prototypes based on the research of the detection method.in order to realize the practical application of rapid non-destructive testing of apple maturity.The main contents and conclusions of this paper are as follows:(1)The relationship between apple maturity and spectral information was explored,the maturity characteristic wavelength was extracted,and a non-destructive testing model based on a machine learning algorithm was established.The random frog algorithm,successive projection algorithm and random frog combined with successive projection algorithm were used to extract the maturity characteristic wavelength,and the least square support vector machine and linear discriminant analysis were used to establish a maturity lossless prediction model.Compared with the linear discriminant analysis model,the least square support vector machine model had better prediction performance,and the classification accuracy of the prediction set was 87.98-92.35%.The independent verification set was used for external verification of the developed model,and the classification accuracy of the verification set was86.94-88.33%.The results show that the selected characteristic wavelength can be used for non-destructive testing of apple maturity,and the developed model can detect apple maturity quickly and accurately in a long time.(2)The appropriate multi-channel spectral sensor was selected according to the characteristic wavelength,and the spectral detection platform based on the sensor was built to verify the application of the sensor in apple maturity detection.According to the characteristic wavelength,a cost-effective 18-channel multispectral sensor system was selected to realize non-destructive testing of apple maturity.The spectral shape features(spectral ratio,spectral difference and normalized spectral intensity difference)were used to preprocess the original spectral data.The least square support vector machine model was established based on the original spectrum and the processed spectrum.Compared with the original spectrum and single processing method,the best prediction accuracy was obtained by the combination of three spectral shape feature parameters,and the classification accuracy of the prediction set was88.46%.The results show that the multi-channel spectral sensor has the ability to realize nondestructive testing of apple maturity.(3)The software and hardware design of the detection equipment was completed based on the multi-channel spectral sensor,and the stability and accuracy of the detection equipment were verified.The multi-channel spectral sensor combined with a halogen tungsten lamp was used as the spectrum acquisition unit,and the raspberry pi 3b + was used as the control unit.the hardware systems of the detection equipment were designed,such as light source drive,spectrum acquisition and storage,user interaction,detection result storage and printing,equipment shape and so on.The real-time control software of equipment was written based on python language.Experiments were designed to verify the detection accuracy and stability of the equipment.The coefficient of variation of the spectrum collected by the sensor at different wavelengths was less than 1.312%,and the classification accuracy of the verification set was84.72%,and the AUC value of each maturity was greater than 0.8972.The results show that the developed testing equipment can accurately realize the non-destructive testing of apple maturity.