Based on the hyperspectral imaging of bacon,the CNN-SVM model designed in this paper organically combines deep learning extraction features with traditional machine learning extraction features to design an accurate and reliable bacon nutrition and health risk four classifier.The three-dimensional convolutional neural network is used to extract the deep features of the hyperspectral image of bacon,and the spectral features of the hyperspectral are fused.Both input the support vector machine (SVM) to realize the classification and health risk assessment of bacon,which is comparable to the national bacon biochemical detection standard.Consistent hyperspectral nutrition quality detection and health risk assessment indicators have achieved the research purpose of reliable and rapid evaluation of its health and nutrition quality.Based on the two classifications of bacon,the accuracy of the four classifications achieved by this method reaches 92.5%.The experimental results prove the feasibility and effectiveness of this method.