%0 Journal Article
%T 融合场景机制的机械作业操作识别算法<br>Mechanical Operation Recognition Algorithm Incorporating Scene Mechanism
%A 宋浪
%J Modeling and Simulation
%P 2339-2346
%@ 2324-870X
%D 2024
%I Hans Publishing
%R 10.12677/mos.2024.133214
%X 机械作业场景下的操作识别可以大大提高安全监督效率，针对传统的目标检测算法无法根据不同的机械作业场景识别操作行为的问题，提出一种融合场景机制的机械作业操作识别算法，融合两个在COCO数据集上预训练好的YOLOv5s模型，先进行机械作业场景识别，再进行关键目标检测，同时利用交并比构建逻辑函数对作业人员的操作行为进行判定。以角磨机作业场景为例，经实验验证，本文的算法模型识别精确率为97.9%，平均识别时间为0.114 s，满足了精确性与实时性要求。<br />
Operation recognition in mechanical operation scenarios can greatly improve the efficiency of safety supervision. Aiming at the problem that traditional target detection algorithms cannot identify operation behaviors according to different mechanical operation scenarios, a mechanical operation recognition algorithm integrating scene mechanism is proposed, which integrates two YOLOv5s models pre-trained on COCO data sets. First, the mechanical operation scene is identified, and then the key target is detected. At the same time, the operation behavior of the operator is judged by the logic function constructed by the intersection ratio. Taking the operation scene of Angle mill as an example, the experimental verification shows that the algorithm model recognition accuracy rate is 97.9%, and the average recognition time is 0.114s, which meets the accuracy and real-time requirements.
%K 深度学习，目标检测，机械作业，YOLOv5s，交并比<br>Deep Learning
%K Object Detection
%K Mechanical Operation
%K YOLOv5s
%K Intersection over Union
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=87004