Research on the Differentiation of Ginseng Radix and Panacis Quinquefolii Slices Based on Convolutional Neural Network

OBJECTIVE To establish an automate method for differentiating slices of Ginseng radix and Panacis quinquefolii based on convolutional deep neural network. METHODS A data set containing the authenticity，of Ginseng radix and Panacis quinquefolii slice was constructed， and the learning rate， momentum coefficient， batch size， and weight attenuation coefficient of the convolutional neural network were optimized by orthogonal experiments to determine the convolutional neural network and the best conditions for Ginseng radix and Panacis quinquefolii. RESULTS A method based on convolutional neural network was established to discriminate and differentiate Ginseng radix slice from Panacis quinquefolii slice. The recognition accuracy rate was 0.909. CONCLUSION This study used convolutional neural network image recognition technology to establish a method for the authenticity identification of Ginseng radix and Panacis quinquefolii slices， which is suitable for rapid differentiation of Ginseng radix and Panacis quinquefolii slice.