计算机应用

• 人工智能与仿真 •    下一篇

施工作业面安全帽的深度学习检测方法

杨静,张育飞,毛晓琦   

  1. 西安理工大学 机械与精密仪器工程学院
  • 收稿日期:2020-04-26 修回日期:2020-07-08 发布日期:2020-07-08 出版日期:2020-08-14
  • 通讯作者: 张育飞

Deep learning detection method of safety helmet on construction working surface

  • Received:2020-04-26 Revised:2020-07-08 Online:2020-07-08 Published:2020-08-14

摘要: 摘 要: 针对施工作业面复杂背景下安全帽检测存在的遮挡,光照多变,且目标尺寸不一等检测问题,提出了一种基于YOLO-v3(You Only Look Once)深度学习网络的改进方法。将原网络模型的52*52检测特征图2倍上采样后,与网络中第二个残差块输出的特征图进行信息融合,再经过一组卷积操作得到一个104*104的新检测特征图。该特征图连同原模型的三个特征级联,构成一个四尺度的目标检测网络模型,改进的检测网络实现了低层特征图的边缘轮廓信息与高层特征图的语义信息的融合。通过实际现场不同工况的测试,结果表明,与原YOLO-v3模型相比,改进模型检测的召回率,准确率以及F1值分别提高了3.7,0.7以及2.29个百分点,满足施工作业面复杂背景下安全帽检测的准确率要求。

Abstract: Abstract: Aiming at the detection problems of helmets in the complex background of construction surface, such as occlusion, changeable illumination and different target sizes, an improved method based on the deep learning network of YOLO-v3 (You Only Look Once) was proposed. After the 52 * 52 detection feature map of the original network model was upsampled twice, it was fused with the feature map output by the second residual block in the network, and then a new detection feature map of 104 * 104 was obtained through a group of convolution operations. It was cascaded with the three feature maps of the original model to form a four scale target detection network model, the improved detection network realized the fusion of the edge profile information of the low-level feature map and the semantic information of the high-level feature map.Through the testing of various working conditions in the actual site, the results show that compared with the original model, the recall rate, accuracy rate and F1 value of the improved model improves by 3.7, 0.7and 2.29 percentage points respectively, the model can meet the requirements of the accuracy of helmet detection in the complex background of construction surface.

中图分类号: