基于YOLOv5的行李运输系统轮对裂纹检测算法
    点此下载全文
引用本文:方浩楠,李登鹏.基于YOLOv5的行李运输系统轮对裂纹检测算法[J].计算技术与自动化,2024,(2):77-81
摘要点击次数: 32
全文下载次数: 0
作者单位
方浩楠,李登鹏 (中国民航大学 电子信息与自动化学院天津 300300) 
中文摘要:为了检测机场行李运输系统中轮对存在的裂纹,提出了一种基于YOLOv5的机场行李运输系统轮对裂纹检测方法。该方法针对裂纹细小、密集等特点采取了以下措施:在YOLOv5网络的 Head 部分使用SIoU替换了原有的CIoU;在Backbone部分加入了SE注意力机制;在Neck部分引入了Swin-Transformer模块;在整个YOLOv5 网络中使用SPD-Conv替代了传统的Conv卷积模块;在图像的预处理方面,使用了图像分割与子图反向拼接技术。通过这些改进,有效地改善了YOLOv5对于细小、密集裂纹的特征提取能力,相较于传统的 YOLOv5 算法,裂纹检测的能力得到了有效提升。
中文关键词:裂纹检测  YOLOv5  SPD-Conv(space-to-depth-Conv)  SIoU  注意力机制  图像预处理
 
Wheelset Crack Detection Algorithm of Baggage Transportation System Based on YOLOv5
Abstract:In order to detect the cracks existing in the wheelset of airport baggage transportation system, a new wheelset crack detection method based on YOLOv5 was proposed. In view of the characteristics of small and dense cracks, the following measures were taken. The original CIoU is replaced by SIoU in the Head part of YOLOv5 network. Then the SE attention mechanism is added to the Backbone part and the Swin-Transformer module is introduced in the Neck part. In addition, The traditional Conv convolution module is replaced by space-to-depth-Conv in the entire YOLOv5 network. In terms of image preprocessing, image segmentation and subgraph reverse stitching technology are used. Through these improvements, the feature extraction ability of YOLOv5 for fine and dense cracks is effectively improved, and compared with the traditional YOLOv5 algorithm, the ability of crack detection has been effectively improved.
keywords:crack detection  YOLOv5  SPD-Conv(space-to-depth-Conv)  SIoU  attention mechanism  image preprocessing
查看全文   查看/发表评论   下载pdf阅读器