YOLO-DCRCF: An Algorithm for Detecting the Wearing of Safety Helmets and Gloves in Power Grid Operation Environments

Area of Science:

• Electrical Engineering
• Computer Vision
• Occupational Safety

Background:

• Personal protective equipment (PPE) like safety helmets and gloves are crucial in power grid operations.
• Traditional detection methods for PPE are often inaccurate due to environmental challenges like poor lighting, occlusions, and dense workforces.

Purpose of the Study:

• To develop a novel algorithm, YOLO-DCRCF, for improving the accuracy and reliability of detecting safety helmet and glove usage in power grid environments.
• To address the limitations of existing detection methods in complex operational settings.

Main Methods:

• Proposed the YOLO-DCRCF algorithm, an enhancement of YOLOv11, integrating Deformable Convolutional Network version 2 (DCNv2) for better geometric transformation modeling.
• Introduced a recalibration feature pyramid (RCF) network to improve feature interaction and multi-scale information capture.

Main Results:

• Achieved a mean Average Precision at 50% IoU (mAP50) of 92.7% on the Safety Helmet Wearing Dataset (SHWD).
• Attained an mAP50 score of 79.6% on the Safety Helmet and Gloves Wearing Dataset (SHAGWD).
• Outperformed the baseline YOLOv11 model by 1.1% on SHWD and 2.7% on SHAGWD.

Conclusions:

• The YOLO-DCRCF model demonstrates superior performance in detecting safety helmets and gloves in power grid environments.
• The proposed algorithm meets the real-time safety monitoring requirements for power grid operations, enhancing worker safety.