Research On Pear Surface Damage Detection Based On Fluorescence Hyperspectral Imaging Technique

Fruit is an indispensable part of people’s daily life,and its quality and safety issues are of great concern.In the process of growing,transporting,and selling fruits,various factors may cause quality problems,among which mechanical damage is a fatal injury to fruits in post-harvest storage.The traditional human observation method can detect the surface damage of fruits,and the application is more mature,but such methods are time-consuming and labor-intensive,and easy to miss/wrong inspection.The fruit industry urgently needs a non-destructive and accurate method to detect early damage to fruits.Currently,hyperspectral imaging technology with both spectral and image information is widely used for nondestructive detection of the quality of food or agricultural products,etc.Fluorescence hyperspectral imaging technology has received more and more attention from researchers in fruit quality detection under its high sensitivity and high specificity.Based on fluorescence hyperspectral imaging technique,this project analyzes the spectral differences of different damage types,different impact degrees,and damage times of pears by obtaining fluorescence spectral information and spatial information for early damage detection of pears,explores and elucidates the fluorescence signal change patterns and causes of pears after damaged by combining microstructural information of damaged tissues.In the study of fluorescence hyperspectral imaging detection of different types of damage of crown pear,the fluorescence effects of healthy fruit and different types of damaged(bruised,scraped,and scratched)fruit of crown pear were investigated.The fluorescence spectral characteristics of different types of fruit were compared and analyzed.And the modeling discrimination analysis of different types of damage was carried out by pre-processing and feature extraction of fluorescence spectral image data.Besides,the microstructure of healthy and different types of damaged pear tissue was analyzed.The results showed that: the fluorescence intensities at 530 nm and 680 nm were significantly different between the healthy pears and the three different types of damaged pears,and the fluorescence intensities at 530 nm were in the order of healthy,bruised,scraped,and scratched;the classification accuracy of SNV-SVM model reached 97.5% for all types of samples;the degree of tissue cell structure damage shown by the electron microscope scan could effectively explain the differences in fluorescence spectra between healthy pears and three different types of damaged fruits.In the study of fluorescence hyperspectral imaging detection of crown pear with different degrees of impact and damage time,the fluorescence spectral characteristics of healthy pear and pear with different degrees of impact(mild,severe)and five different moments(immediate,15 minutes,1 day,2 days and 3 days)after damage were investigated.The modeling and discriminative analysis of the pear flesh sections were carried out by pre-processing and feature extraction,and the microstructure of the pear flesh sections was analyzed.The results showed that: the fluorescence signal of crown pear(especially at 530 nm and 680 nm peaks)gradually weakened with the increasing degree of damage,while with the extension of damage time,the fluorescence signal of healthy parts remained unchanged but the fluorescence intensity of the damaged parts(both severely and lightly damaged)gradually decayed;the SVM model could identify the different degrees of damage with an accuracy of 93.33% after 15 minutes,and the accuracy increases to 96.67% after 48 hours;for samples with different damage times,the accuracy is up to 90.00% under the degree of severe impact;the overall differentiation accuracy between damaged and healthy samples can reach 99.33%;the microstructure can also effectively elucidate the fluorescence spectral change patterns of different samples.The results showed that the fluorescence hyperspectral imaging technique can be used not only for the discrimination analysis of different damage types of crown pear but also for the discrimination analysis of different impact degrees and damage time.This research has important significance and application value to the development of fruit quality nondestructive testing technology.