An Improved Fast Detection Method for Helmets in Complex Construction Environment
- Resource Type
- Conference
- Authors
- Fan, Haifeng; Jiang, Kun; Hu, Rongyi; Wang, Tingting; Dong, Long; Zhu, Siyu
- Source
- 2022 4th International Conference on Machine Learning, Big Data and Business Intelligence (MLBDBI) MLBDBI Machine Learning, Big Data and Business Intelligence (MLBDBI), 2022 4th International Conference on. :132-136 Oct, 2022
- Subject
- Computing and Processing
Head
Convolution
Object detection
Machine learning
Feature extraction
Data models
Real-time systems
Lightweight
YOLOv4
target detection
MobileNetV3
K-means++
- Language
To solve the problem that the detection speed of the existing helmet detection algorithm cannot meet the requirements of the practical application problem, we propose a lightweight helmet detection algorithm based on YOLOv4 as the main framework. The algorithm optimizes the backbone network and feature pyramid of YOLOv4. It replaces the original backbone network of YOLOv4 with MobileNetv3 which can improve the computational efficiency by depthwise separable convolution. The algorithm also adds a spatial pyramid module to the feature pyramid part which can improve the detection accuracy of the algorithm. Finally, we obtain the anchor boxes through the K- means++ algorithm to accelerate the convergence speed of the algorithm and improve the accuracy of the model. The experimental results show that the mAP value is 90.4%, and the detection speed can reach 63f/s on a single card RTX 3090.