Classification and detection of insects from field images using deep learning for smart pest management: A systematic review

• Published in: Ecological informatics Vol. 66; p. 101460
• Main Authors: Li, Wenyong, Zheng, Tengfei, Yang, Zhankui, Li, Ming, Sun, Chuanheng, Yang, Xinting
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 01.12.2021
• Subjects: Classification and detection; Computer vision; crop yield; crops; Deep learning; insect pests; Plant protection; prediction; Smart pest monitoring; systematic review; Deep learning; Computer vision; Plant protection; Smart pest monitoring; Classification and detection
• ISSN: 1574-9541
• DOI: 10.1016/j.ecoinf.2021.101460

Insect pest is one of the main causes affecting agricultural crop yield and quality all over the world. Rapid and reliable insect pest monitoring plays a crucial role in population prediction and control actions. The great breakthrough of deep learning (DL) technology has resulted in its successful applications in various fields, including automatic insect pest monitoring. DL creates both new strengths and a series of challenges for data processing in smart pest monitoring (SPM). This review outlines the technical methods of DL frameworks and their applications in SPM with emphasis on insect pest classification and detection using field images. The methodologies and technical details evolved in insect pest classification and detection using DL are summarized and distilled during different processing stages: image acquisition, data preprocessing and modeling techniques. Finally, a general framework is provided to facilitate the smart insect monitoring and future challenges and trends are highlighted. In a word, our purpose is to provide researchers and technicians with a better understanding of DL techniques and their state-of-art achievements in SPM, which can promote the implement of various SPM applications. •Applications of deep learning (DL) in smart pest monitoring (SPM) were investigated comprehensively.•Methodologies involved in were analyzed and distilled from these applications.•A general framework was synthesized to guide the applications of DL in SPM.•Challenges and possible solutions of DL in SPM were discussed.