@Article{cmc.2026.087447,
AUTHOR = {Shuai Zhou, Jun Wu, Li Li, Rong Tang, Zhangjun Peng, Mingfei Wan, Zhiqiang Chen, Zhigui Liu},
TITLE = {Intelligent Crop Pest Identification: From Traditional Machine Learning to Multimodal Large Models—A Review},
JOURNAL = {Computers, Materials \& Continua},
VOLUME = {},
YEAR = {},
NUMBER = {},
PAGES = {{pages}},
URL = {http://www.techscience.com/cmc/online/detail/28331},
ISSN = {1546-2226},
ABSTRACT = {Crop pests pose a major threat to agricultural production and hinder the sustainable development of agriculture. Accurate pest identification is a prerequisite for implementing sustainable pest control strategies and enabling precise and efficient pest management. This review examines the technological evolution of intelligent crop pest identification and analyzes the advantages and limitations of existing methods. First, it analyzes three types of crop pest data collection methods, with emphasis on sampling bias, existing bottlenecks such as long-tail distribution and domain shift, and the characteristics of typical datasets. Second, it discusses the application scenarios and limitations of traditional machine learning methods, including clustering, support vector machines, and decision trees. Next, it compares the technical characteristics and performance differences of one-stage and two-stage convolutional neural networks, as well as Vision Transformer models, in pest identification. Then, it discusses the decoupled and end-to-end architectures of multimodal large models and their cross-modal applications in intelligent pest diagnosis, thereby offering new perspectives for closed-loop pest diagnosis and ecologically based prevention and control decision-making. Finally, this review summarizes the limitations of existing research and proposes future research directions for developing more reliable and efficient crop pest identification technologies.},
DOI = {10.32604/cmc.2026.087447}
}