| The detection and prevention of insect pests in stored grain is the focus of grain storage industry. Effective detection and disturbance is of great significance. In view of the defects of existing detection of concealment pest, the paper combines pattern recognition with biological photon technology to detect insect pests. The focus is to study ultra weak luminescence signal measurement analysis and pattern classification algorithm design of normal wheat and wheat with hidden insect.The research content and innovation include:(1)The insect density and larval stage as the research object, the actual out rate based on different dye insect density is studied. Combining density with classification model, two kinds of insect density samples which the next generation insect rate closes to saturation are selected as analysis object, and larval(three ages and four ages) as study object..(2)After the noise is reducted, sensitive feature of ultra weak luminescence signal is extracted.The eight statistical characteristics and twelve histogram features are extracted as characteristic vector of grain. Then principal component analysis is performed.(3)Based on the characteristics of small sample data, linear discriminant analysis(LDA)and quadratic discriminant analysis(QDA) are used to realize the model classification of nomal wheat and wheat with hidden insect. For limitations of QDA, regularization algorithm is used to solve.(4)Support vector machine(SVM) algorithm is used to design classifier. In order to achieve the best performance, grid search algorithm is used for related parameters optimization to achieve global optimal. Finally, three recognition rates of the algorithm are compared, and the optimal classifier is chose.Combineing biological ultra weak luminescence mechanism with pattern recognition technology, it is feasible to test the larval pest. the research results lay the foundation for the further study of detection. |
PDF Full Text Request