With the continuous development and expansion of the Chinese herbal medicine market, the quality control of herbs has become particularly important. In this study, a micro-morphological feature recognition method based on deep convolutional neural network (CNN) for the traditional Chinese herbal medicine maitake is proposed. First, images of different grades of maitake were collected, and multiple features such as morphology, color, and texture were extracted after image preprocessing, feature extraction, and enhancement. Then, the extracted images were classified and recognized using a deep learning algorithm. The experimental results show that the top-1 accuracy of this paper’s method on the test set is 97.14%, and the top-1 accuracy after migration learning is improved to 99.35%, and the macro accuracy reaches 99.54%. Compared with traditional algorithms combining image processing and machine learning, the method in this paper has significant advantages. In addition, the depthseparable convolutional structure effectively reduces the computational burden of the model. The method shows good application prospects in the quality identification of Chinese herbal medicine Maitong, and can provide powerful technical support for the quality control of Chinese herbal medicine.
