Design And Implementation Of A Rice Classification Algorithm Based On Fusion Features

In recent years,there are many types of rice in the market,and it is difficult for consumers to distinguish the authenticity of the type.Therefore,a fast and effective method of rice classification is needed to help consumers determine the type of rice.As traditional detection methods,sensory and chemical detection are complicated and lack timeliness.The use of machine vision technology to classify rice images of different qualities is not only fast and can handle a large amount of information,but it is also more stable and accurate than traditional methods.This paper studies the color and texture feature extraction and classification algorithms of five types of rice(Brown Rice,Engineering North Rice,Golden Germ Rice,Polished Rice,Embryo Rice).The five types of rice have different surface texture characteristics due to their different processing methods.This paper classifies rice based on LBP texture characteristics.However,the texture features extracted by the traditional LBP algorithm have shortcomings such as large dimensions and data redundancy.Therefore,this paper proposes a PCA optimized dimensionality reduction LBP feature extraction method.On the premise of retaining 98% of the original feature information,the traditional LBP algorithm extracts 256 The dimensional feature is reduced to 4 dimensions.Through comparative experiments with the three encoding modes of LBP(original LBP,equivalent LBP,rotation invariant equivalent LBP),it is verified that the method proposed in this paper has lower computational complexity while ensuring the classification performance.In addition to texture features,the average value of G and B channels in the RGB color space of rice images can also be used as classification features of rice to distinguish part of rice.Therefore,this paper proposes a multi-feature fusion algorithm based on color features and texture features.The four-dimensional features extracted by the PCA optimized dimension reduction LBP feature extraction method and the G and B channel color average features are combined to form a six-dimensional feature vector,and the fusion is The feature vector is used as the input of the classification feature to realize the classification of rice.Based on the fusion features,this paper uses a support vector machine to design a rice classifier.Through the comparison experiment of linear kernel,polynomial kernel and Gaussian kernel,the Gaussian kernel with high classification accuracy is selected as the kernel function of the classifier.K-fold cross-validation is used to optimize the parameters of the classifier.It is verified by comparative experiments that the fusion multi-feature extracted by the method in this paper is more suitable for the classification of five types of rice,and the classification accuracy rate reaches 95.1% under the condition of lower computational complexity.