A Visual and Automated Calculation Method for Hazardous Areas of Live-Line Work Based on Point Cloud

Traditional live-line work often relies on experience and manual labor. At present, the safety assessment of live-line work scenes adopts unified standards; however, due to the complexity of live-line work scenes, the unified standard cannot accurately measure and evaluate its safety. In order to address these issues, this article proposes a new method for identifying the hazardous areas in live-line work scenes. The proposed method consists of 2 parts. First, PointMGA, whose baseline model is PointNet++, is proposed to segment the point clouds of the transmission scenes. The experimental results demonstrate that the PointMGA achieves higher segmentation accuracy compared to both the PointNet++ model and PointNet segmentation models. Specifically, our method achieved an overall accuracy (OA) of 99.1% and 99.5% on the validation and testing sets, respectively, with a Mean Intersection over Union (MIoU) of 96.7% and 99.7%. The second part of the method involves using the Farthest Point Sampling-Bisection Region Growing (FPS-BRG) distance algorithm to calculate the hazardous areas within the specific transmission scenes. This algorithm ensures accuracy and significantly improves computational efficiency by over eight times. The results demonstrate the effectiveness of the approach in enhancing the automation and safety of live-line work in complex transmission scenes.