Key Technology Research Of Content-based Chinese Herbal Medicine Image Search

Content-based image retrieval of Chinese herbal plants is an application of content-based image retrieval technology in the field of Chinese herbal plants. The purpose of it is to search similar images in the image database of Chinese herbal plants according to some expressive visual features. However, different from the traditional content-based image retrieval, it’s more concerned about the fine-grained image retrieval and more challenging. In this paper, we do some research on the content-based image retrieval of Chinese herbal plants. The primary task and contribution of this paper are as follows.1. We propose a web image crawler based on the professional word vector in the field of traditional Chinese medicine and construct an image database of Chinese herbal plants.2. We conduct a thorough study on bag-of-words model. First of all, we assess the impact of patch size and dictionary size; In the next place, we study and compare different encoding methods, including hard-histogram encoding, soft-histogram encoding, locality-constrained linear encoding, Fisher vector encoding and super vector encoding; Once more, we analysis two pooling methods-max pooling and average pooling. Last, we propose a method combing SIFT and SIFT-based color descriptors for efficient image retrieval of Chinese herbal plants.3. We study the architecture of AlexNet and GoogLeNet model and then build corresponding convolution neural network according to their architecture.4. We design and implement an image retrieval system of Chinese herbal plant, including image feature extraction, feature indexing and image retrieval.