农业机械学报

赵辉,黄镖,王红君,岳有军.基于改进YOLO v7的农田复杂环境下害虫识别算法研究[J].农业机械学报,2023,54(10):246-254.
ZHAO Hui,HUANG Biao,WANG Hongjun,YUE Youjun.Pest Identification Method in Complex Farmland Environment Based on Improved YOLO v7[J].Transactions of the Chinese Society for Agricultural Machinery,2023,54(10):246-254.

摘要点击次数: 258
全文下载次数: 143

基于改进YOLO v7的农田复杂环境下害虫识别算法研究 [下载全文]

Pest Identification Method in Complex Farmland Environment Based on Improved YOLO v7 [Download Pdf][in English]

投稿时间：2023-04-11

DOI：10.6041/j.issn.1000-1298.2023.10.024

中文关键词: 农田复杂环境害虫识别目标检测深度学习 YOLO v7 Swin Transformer

基金项目:天津市科技支撑计划项目（19YFZCSN00360）

作者	单位
赵辉	天津理工大学
黄镖	天津理工大学
王红君	天津理工大学
岳有军	天津理工大学

中文摘要:为使巡检机器人能够对体积小且密集、形态多变、数量多且分布不均的害虫进行高效精准识别，提出了一种基于改进YOLO v7的害虫识别方法。该方法将CSP Bottleneck与基于移位窗口Transformer（Swin Transformer）自注意力机制相结合，提高了模型获取密集害虫目标位置信息的能力；在路径聚合部分增加第4检测支路，提高模型对小目标的检测性能；将卷积注意力模块（CBAM）集成到YOLO v7模型中，使模型更加关注害虫区域，抑制背景等一般特征信息，提高被遮挡害虫的识别精确率；使用 Focal EIoU Loss 损失函数减少正负样本不平衡对检测结果的影响，提高识别精度。采用基于实际农田环境建立的数据集的实验结果表明，改进后算法的精确率、召回率及平均精度均值分别为91.6%、82.9%和88.2%，较原模型提升2.5、1.2、3个百分点。与其它主流模型的对比实验结果表明，本文方法对害虫的实际检测效果更优，对解决农田复杂环境下害虫的精准识别问题具有参考价值。

ZHAO Hui HUANG Biao WANG Hongjun YUE Youjun

Tianjin University of Technology

Key Words:complex farmland environment pest identification target detection deep learning YOLO v7 Swin Transformer

Abstract:In order to enable the inspection robot to efficiently and accurately identify small, dense, morphologically variable, numerous and unevenly distributed pests, a pest recognition method based on the improved YOLO v7 was proposed. CSP Bottleneck was combined with a self attentional mechanism based on shift window transformer (Swin Transformer), which improved the ability of the model to obtain the location information of dense pests. A fourth detection branch was added to the path aggregation part to improve the detection performance of the model on small targets. The convolutional attention module (CBAM) was integrated into the YOLO v7 model to make the model pay more attention to the pest area, suppress the background and other general feature information, and improve the recognition accuracy of blocked pests. Focal EIoU Loss function was used to reduce the influence of positive and negative sample imbalance on detection results and improve the recognition accuracy. According to the experimental results, the accuracy rate, recall rate and mAP of the improved algorithm were 91.6%, 82.9% and 88.2%, respectively by using the data set established based on the actual farmland environment, which was 2.5, 1.2 and 3 percentage points higher than that of the original model. Compared with other mainstream models, the experimental results showed that the proposed method was more effective in the actual detection of pests, and it had practical application value in solving the problem of accurate identification of pests in complex farmland environment.

Transactions of the Chinese Society for Agriculture Machinery (CSAM), in charged of China Association for Science and Technology (CAST), sponsored by CSAM and Chinese Academy of Agricultural Mechanization Science(CAAMS), started publication in 1957. It is the earliest interdisciplinary journal in Chinese which combines agricultural and engineering. It always closely grasps the development direction of agriculture engineering disciplines and the published papers represent the highest academic level of agriculture engineering in China. Currently, nearly 8,000 papers have been already published. There are around 3,000 papers contributed to the journal each year, but only around 600 of them will be accepted. Transactions of CSAM focuses on a wide range of agricultural machinery, irrigation, electronics, robotics, agro-products engineering, biological energy, agricultural structures and environment and more. Subjects in Transactions of the CSAM have been embodied by many internationally well-known index systems, such as: EI Compendex, CA, CSA, etc.

下载PDF阅读器