Research On Tea Appearance Quality Classification Detection Technology Based On Multi-Feature Fusion

As a major tea producer and consumer,it is particularly important to control the quality of tea from the perspective of satisfying consumer needs or the perspective of the long-term development of the enterprise.At present,the tea industry generally uses sensory evaluation,physical and chemical indicators,mechanical grading,and other testing methods to judge the quality of tea.According to the tea grading standards,the appearance and sensory quality characteristics are the most direct reference direction for grading detection.Because of the current stage,manual detection is the most common sensory detection method,but manual detection is highly subjective,requires high professional reviewers,and has complicated grading procedures;physical and chemical identification needs to destroy the shape of the tea,and it is mostly used for the determination of the contents of the tea.Chemical experimental research;while mechanical grading is mostly used in the picking stage to classify fresh tea leaves;of course,there are also new technologies that are used in tea quality evaluation,such as electronic nose,electronic tongue,fluorescence spectrum evaluation technology,etc.;for new technologies Although the application in the appearance evaluation of tea can make up for the shortcomings of sensory evaluation and physical and chemical testing methods,it mostly focuses on the identification of tea types,the analysis of internal chemical composition content,and the quality evaluation of the single visual characteristics of tea.The classification result is not accurate enough.To this end,because of the current subjectivity of manual grading of tea,the defects of mechanical damage to tea in assembly line work,and the single-sidedness of the current appearance grading method,a multi-feature fusion based on evidence theory is proposed.Tea grading method.This article takes a sample of Buxus from Yibin,Sichuan as the research object,and does the following:(1)Select experimental hardware equipment,build an experimental sample collection platform,and perform data preprocessing on the collected images;(2)To improve the accuracy of classification detection and avoid the one-sidedness of single feature classification,the color,and shape of tea leaves are analyzed based on computer vision technology,combined with impurity characteristics,and the related algorithms for feature extraction are studied;color moments are used as the characteristics of tea leaves.Color features;Hu invariant moments are used as shape parameters;for tea impurity detection,an improved Hough transform algorithm is used to detect the impurity characteristics of tea;(3)To avoid the problem of excessive computation caused by the fusion of multiple data features in the tea detection process,principal component analysis(PCA)is used to reduce the dimensionality and determine the feature parameters with a large contribution rate.Combining the extracted tea features,the support vector machine(SVM)is used to initially classify a single feature of the tea,and the idea of multi-feature fusion is introduced.The preliminary single feature classification result output by the support vector machine is used as evidence and used in the future.The DS theory of ability and advantage in terms of certainty and feature combination,finalize the grading of tea,and obtain the result of multi-feature fusion grading of tea;Through simulation comparison,the accuracy rate of SVM single feature classification result is 82%-86%,and the accuracy rate of SVM multi-feature classification result is about93%.Using SVM single feature classification result as evidence,the multi-feature fusion classification obtained by DS evidence theory is accurate The rate is over 96%.Experiments show that the introduction of evidence theory of multi-feature fusion tea grading method further improves the reliability of tea quality grading detection and the correct rate of grading.It is an effective automatic tea grading method,which can improve the non-destructive,Efficient and intelligent have laid the foundation,making the research results have industrialization significance.