《环境昆虫学报》

物种种类识别与计数是田间昆虫目标检测的重要内容，其对害虫的监测预警及科学防控具有重要意义。传统人工识别昆虫种类和计数的方法效率低，难以应对田间昆虫种类的多样性，且无法满足智慧化农业对害虫防控工作的需要。随着计算机与互联网技术的快速发展，昆虫目标检测手段逐渐智能化、精准化。基于图像的昆虫目标检测方法凭借其高效、易操作、适用范围广等优势，成为近年来国内外昆虫种类识别与计数研究热点和主要技术手段。本文综述了传统目标检测算法特征提取技术和分类器；详述了基于锚框（Anchor based）深度学习目标检测模型，如YOLO（You only look once）系列、SSD（Single shot multibox detector）系列等；介绍了无锚框（Anchor free）深度学习目标检测模型，如CornerNet系列等。本文还探讨了基于图像的昆虫目标检测存在的问题及未来的研究方向。

英文摘要:

Species identification and counting is an important content of target detection of insects in the field, which is of great significance to the monitoring, early warning and scientific prevention and control of pests. Traditional methods for identifying and counting insect species were inefficient and struggled to address the diverse range of insects encountered in the field, falling short of the demands of intelligent agriculture for effective pest management. However, with the rapid development of computer and internet technology, insect target detection methods have evolved to become increasingly intelligent and precise. In recent years, image-based insect target detection has become the primary technical approach for insect species identification and counting both domestically and internationally. This paper reviews the feature extraction techniques and classifiers of traditional target detection algorithms. Furthermore, it describes anchor based deep learning object detection models, such as YOLO (You only look once) series, SSD (Single shot multibox detector) series. Additionally, the paper introduces anchor free deep learning object detection models, such as the CornerNet series. Lastly, it discusses the prevalent challenges and future research directions in the realm of image-based insect target detection.

View Full Text View/Add Comment Download reader