Fine-grained image classification of cashmere wool based on sparse dictionary learning

Abstract

The identification of cashmere and wool fibers has always been a key issue in the wool textile industry. Cashmere and wool fibers are similar in morphology, and the morphological changes of similar fibers are diverse, and the differences within the category are subtle. The appearance and morphology of cashmere and wool show reverse changes mainly as follows: the cashmere becomes thicker and the wool fiber becomes thinner. Affected by changes in breeding methods, breed improvement and climate change, cashmere wool fibers have more and more variant forms, which have a certain impact on the subsequent production and processing of cashmere wool fibers. First, we preprocess the fiber image to achieve data enhancement and acquire key regions. Then, through sparse dictionary learning, image related information such as dictionary and sparse coding is obtained, and finally, the related information is used to realize the fine-grained image classification of cashmere wool. The experimental results show that this method has better classification effect than other methods.