%0 Journal Article
%T YOLOv5改进算法在机械零件中的识别与应用<br>Identification and Application of YOLOv5 Improved Algorithm in Mechanical Parts
%A 张浩洋
%A 何仕荣
%A 孟冬平
%J Software Engineering and Applications
%P 1446-1455
%@ 2325-2278
%D 2022
%I Hans Publishing
%R 10.12677/SEA.2022.116149
%X 针对智能生产制造过程中，传统的目标检测算法对机械零件识别率不高，识别速度慢等问题，结合深度学习与现有算法，提出一种基于YOLOv5改进的目标检测算法。该算法在YOLOv5骨干网络中加入注意力机制，以用来改善原始YOLOv5算法对相似背景下相似零件识别率低的问题；其次，通过引入CIou损失函数，使得改进后的YOLOv5算法能够更快地收敛并具有更好的性能；最后，将改进前后的算法模型分别训练后对机械零件进行识别并对比分析，证实改进后的YOLOv5算法具有高的识别精度和鲁棒性。<br />
Aiming at the problems such as low recognition rate and slow recognition speed of mechanical parts in the process of intelligent production and manufacturing, an improved object detection algorithm based on YOLOv5 was proposed by combining deep learning and existing algorithms. This algorithm adds an attention mechanism into the YOLOv5 backbone network to improve the low recognition rate of similar parts under similar backgrounds of the original YOLOv5 algorithm; Secondly, by introducing CIou loss function, the improved YOLOv5 algorithm can converge faster and have better performance; Finally, the algorithm models before and after the improvement are trained respectively to identify the mechanical parts and make a comparative analysis. It is confirmed that the improved YOLOv5 algorithm has high recognition accuracy and robustness.
%K 深度学习，骨干网络，YOLOv5，机械零件识别<br>Deep Learning
%K Backbone Network
%K YOLOv5
%K Machine Part Identification
%U http://www.hanspub.org/journal/PaperInformation.aspx?PaperID=59942