Ellipsoid-Like Herbaceous Tibetan Medicinal Materials Recognition Using DenseNet with Attention and LBP-Encoded Gabor Features

Tibetan medicine materials play an important role in Tibetan culture. Some types of Tibetan medicinal materials are similar in shape and color, but have completely different medicinal properties and functions. Wrong use of such medicinal materials may produce poisonousness, delay treatment, and bring incalculable consequences to patients. Existing recognition methods of medicinal materials with similar shapes and colors mainly rely on manual appraisal through observation, touch, taste, and smell. In this paper, we propose a deep learning based method for recognizing ellipsoid-like herbaceous Tibetan medicinal materials. We establish an image dataset of ellipsoid-like herbaceous Tibetan medicinal materials, which contains 3200 images of 18 types of ellipsoid-like Tibetan medicinal materials. These medicinal materials are highly similar in shape and color with each other. To cope with the complex background and highly similar shape and color of the ellipsoid-like herbaceous Tibetan medicinal materials in the images, we use DenseNet with Attention and LBP-Encoded Gabor Features to recognize the ellipsoid-like herbaceous Tibetan medicinal materials images. The experimental results show that our method can achieve 92.38% accuracy for the 18 types of Ellipsoid-like herbaceous Tibetan medicinal materials and 90.13% accuracy for two highly similar medicinal materials.