Automatic Safety Helmet Wearing Detection

Kang Li, Xiaoguang Zhao, Jiang Bian, Min Tan

Abstract

Surveillance is essential for the safety of power substation. The detection of whether wearing safety helmets or not for perambulatory workers is the key component of overall intelligent surveillance system in power substation. In this paper, a novel and practical safety helmet detection framework based on computer vision, machine learning and image processing is proposed. In order to ascertain motion objects in power substation, the ViBe background modelling algorithm is employed. Moreover, based on the result of motion objects segmentation, real-time human classification framework C4 is applied to locate pedestrian in power substation accurately and quickly. Finally, according to the result of pedestrian detection, the safety helmet wearing detection is implemented using the head location, the color space transformation and the color feature discrimination. Extensive compelling experimental results in power substation illustrate the efficiency and effectiveness of the proposed framework.

Introduction

Power substations are high-risk environments where strict adherence to safety protocols, such as wearing safety helmets, is critical. While automated surveillance has advanced in monitoring equipment health, ensuring worker compliance has traditionally remained a manual, labor-intensive task. This paper presents a practical computer vision pipeline designed to automate helmet detection in these complex industrial settings.

The Proposed Framework

To balance accuracy with real-time performance, the researchers developed a three-stage cascaded framework. This approach minimizes computational overhead by progressively narrowing the search space from the entire video frame down to the specific head region of a worker.

The three-stage cascaded framework for safety helmet detection.

1. Motion Segmentation (ViBe)

The system first employs the Visual Background Extractor (ViBe) algorithm. Instead of analyzing every pixel, ViBe identifies moving objects by comparing current frames to a dynamically updated background model. This step effectively filters out static background clutter like wires and poles, focusing subsequent processing only on areas of interest.

2. Pedestrian Classification (C4)

Once motion is detected, the C4 framework is used to determine if the moving object is a human. C4 utilizes CENTRIST (Census Transform histogram) features, which prioritize contour cues. This makes the system more robust in metallic environments where the texture of a worker's clothing might blend into the background.

3. Helmet Detection

For each identified pedestrian, the system isolates the head region to verify helmet usage. This is achieved through a spatial heuristic: the top 20% (1/5th) of the pedestrian's bounding box is designated as the head area.

Head region localization using the top 1/5th of the pedestrian bounding box.

Within this sub-image, the system converts the color data from RGB to the HSV (Hue, Saturation, Value) color space. HSV is more resilient to varying lighting conditions and shadows common in outdoor substations. By applying the Otsu method to the Saturation channel, the system can segment and identify the specific color signature of a safety helmet.

Experimental Results

The framework was evaluated using the standard INRIA Person Dataset for initial training and a custom dataset of 10 videos recorded in actual power substations.

  • Pedestrian Classification: Achieved a mean accuracy of 84.2%.
  • Overall System Performance: The full ViBe+C4+CFD framework reached an Area Under the Curve (AUC) of 94.13%.
  • Comparative Analysis: The proposed method outperformed the traditional ViBe+HOG+SVM approach, which yielded an AUC of 89.20%, representing a significant improvement in detection reliability.

Conclusion and Future Work

The proposed framework demonstrates a robust and efficient solution for industrial safety monitoring. By combining motion segmentation with contour-based human detection and HSV-based color analysis, the system achieves high accuracy even in cluttered environments. Future work will focus on improving detection for small, distant targets where contour features are less distinct and refining the head localization logic to handle more varied pedestrian poses.

Summary updated

Abstract

Surveillance is essential for the safety of power substation. The detection of whether wearing safety helmets or not for perambulatory workers is the key component of overall intelligent surveillance system in power substation. In this paper, a novel and practical safety helmet detection framework based on computer vision, machine learning and image processing is proposed. In order to ascertain motion objects in power substation, the ViBe background modelling algorithm is employed. Moreover, based on the result of motion objects segmentation, real-time human classification framework C4 is applied to locate pedestrian in power substation accurately and quickly. Finally, according to the result of pedestrian detection, the safety helmet wearing detection is implemented using the head location, the color space transformation and the color feature discrimination. Extensive compelling experimental results in power substation illustrate the efficiency and effectiveness of the proposed framework.
BibTeX
@misc{li2018automatic,
  author = {Kang Li and Xiaoguang Zhao and Jiang Bian and Min Tan},
  title = {{Automatic Safety Helmet Wearing Detection}},
  year = {2018},
  eprint = {1802.00264},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1802.00264},
}
Figure