Abstract: Terrestrial arthropods are the most diverse animal taxa and play critical roles in human society and natural ecosystems. However, the taxonomic identification of terrestrial arthropods is a great challenge for diversity survey. With the development of computer vision technology, deep learning image recognition technology has shown a great potential for animal and plant recognition and classification and has increasingly become a new method for animal classification. Compared with plants and vertebrates, terrestrial arthropods were less explored the deep learning algorithm technology for the identification. Based on the diversity and ecological functions of terrestrial arthropods, we systematically introduce the principles, basic types and factors affecting the recognition accuracy of deep learning image recognition technology. Furthermore, we summarized the application cases and problems of deep learning recognition technology in the identification of soil arthropods. We prospected future research priorities in arthropod identification using deep learning, which outlines a milestone for achieving rapid, accurate and optimized identification of terrestrial arthropods, especially soil arthropods.

Key words: terrestrial arthropods, soil fauna, image recognition, deep learning