Image Recognition of Chinese Herbal Medicine Using Adaptive Gamma Correction Based on Convolutional Neural Network

Chinese herbal medicine (CHM) has gained significant attention due to its potential in treating various diseases. CHM image classification is crucial in fields such as smart medical care and medical teaching. However, accurately classifying CHM images has been a challenging task due to their fine-grained variability. To achieve automatic identification and classification of Chinese herbal medicines, we designed a method that combines image processing and deep learning. We established a dataset of CHM images, which includes 10 kinds of common Chinese herbal medicines, with 15,719 images in total. The dataset covers medicinal materials in various natural backgrounds and provides ample samples for training deep learning models. Regarding image processing, we proposed a method based on adaptive gamma correction(AGC) and compared it with CLAHE(Contrast Limited Adaptive Histogram Equalization)and traditional gamma correction(TGC) methods. Subsequently, we employed three pre-training models of convolutional neural networks based on transfer learning, namely GoogLeNet, ResNet-18, and MobileNetV2, to identify CHM images. The results of the verification indicate that the AGC-CNN model outperforms the CLAHE-CNN and TGC-CNN models in terms of classification accuracy and image quality improvement.