Application of machine learning in automatic image identification of insects - a review
Yuanyi Gao, Xiaobao Xue, Guo‐Qing Qin, Kai Li, Jiahao Liu, Yulong Zhang, Xin-Jiang Li
Abstract
Fast and reliable identification of insect species is crucial for pest management, animal quarantine, and effective utilization of insect resources. Traditional morphological classification is time-consuming and laborious, while automatic image identification techniques based on machine learning (ML) can greatly improve efficiency. ML is a promising approach for the automatic image identification, including traditional machine learning (TML) and deep learning (DL). This review outlines the process of automatic image identification of insect based on TML/DL. We highlighted methods of image acquisition, preprocessing, segmentation, identification, and detection. The applications of automatic image identification based on ML of various insect orders are summarized and discussed, with a focus on Coleoptera, Lepidoptera, Hymenoptera, Diptera, and Orthoptera. In the future, researchers can conduct studies in the following aspects, such as constructing reliable public big data sets, minimizing the subjective impact of photography, delving into interpretable DL, and increasing the study of diverse insect species. This review provides a new idea for the development of automatic insect identification, to intervene in the occurrence of pests as soon as possible. This can not only reduce chemical pollution but also contribute to the green earth.