Research On Nondestructive Detection Quality Of Kiwifruit Based On Hyperspectral Imaging Technology

Kiwifruit plays an important role in people’s daily life.Not only does it have a mixture of various fruit flavors,but also has specific value in nutriology and diet therapy,which aids their health.The manual grading method is mostly adopted for the kiwifruit postpartum grading,which is subject to large human and low efficiency.In addition,the kiwifruit quality assessment uses instrumental analysis,and all samples are damaged even wasted.To adress these above-mentioned problems,this paper used hyperspectral imaging technology and chemometric methods to detect the shape characteristics,texture characteristics,and sugar content of kiwifruit,and the model was optimized.’Jinkui’ kiwifruit was used as samples for detecting and classifying the normal and malformed fruits.The results showed that LS-SVM has the best effect on normal fruit,and BPNN had the best effect on abnormal fruit.Comprehensively considered,the best classification model was BPNN,with an overall accuracy rate of 97.56%and two misjudged samples.A feasibility study on the texture characteristics of ’Fenghuang-1’ kiwifruit was conducted.The results showed that the hardness1(H1),chewiness,resilience,peel hardness(PH),average hardness(AH),corrected hardness(CH)models had good prediction performance.The correlation coefficient(R)values were exceeded 0.9 with a low root mean square error,and RPDs were exceeded 2.The shear force(SF)and average shear force(ASF)models had a low accuracy with the RPDs of 1.5-2.0,which can roughly predict the quality.Taking ’Jinkui’ and ’Fenghuang-1’ kiwifruit as samples,the optimal sugar content models were established.Meantime,a simplified sugar content model was built using the common characteristic wavelengths(CCWs)from the optimal model.It is feasible that the simplified model predicted the sugar content of multiple varieties fruits.In’Fenghuang-1’ kiwifruit,the optimal sugar content model was SPA-PLS with an RP of 0.9443,RMSEP of 1.1839,and an RPD of 2.84.In ’Jinkui’ kiwifruit,the optimal sugar content model was SPA-PLS with an RP of 0.9293,RMSEP of 1.0105 and an RPD of 2.56.In the verification result of the simplified model,the R values of the three varieties kiwifruits were between 0.81 and 0.89,the RMSEP between 1.50 and 1.67.The model achieved strong applicability and realized the detection of sugar content of multiple varieties kiwifruit by one model.The above researchs show that the application of hyperspectral imaging technology can be used for nondestructive detection of kiwifruit shape characteristics,and prediction of kiwifruit texture characteristics and sugar content.Meantime,the simplified model provides theoretical support for a single model detecting the sugar content of multiple varieties of fruit.