PointNet++ network framework.
<div><p>As the global economy continues to expand and energy demand increases, the size of power transmission networks continues to grow, making the safety monitoring of transmission towers increasingly important. To address the accuracy deficiencies of existing technologies in predictin...
Saved in:
| Main Author: | |
|---|---|
| Other Authors: | , , , , |
| Published: |
2025
|
| Subjects: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | <div><p>As the global economy continues to expand and energy demand increases, the size of power transmission networks continues to grow, making the safety monitoring of transmission towers increasingly important. To address the accuracy deficiencies of existing technologies in predicting external damage risks to transmission towers, this study proposes a real-time spatial distance measurement method based on monocular vision. The method first uses a Transformer network to optimize the distribution of pseudo point clouds and designs a 3D monocular vision distance measurement method based on LiDAR. Through validation on the KITTI 3D object detection dataset, the method achieved an average detection accuracy increase of 10.71% in easy scenarios and 2.18% to 7.85% in difficult scenarios compared to other methods. In addition, this study introduced a foreground target depth optimization method based on a 2D target detector and geometric constraints, which further improved the accuracy of 3D target detection. The innovation of the study is the optimization of the pseudo point cloud distribution using the transformer network, which effectively captured the global dependencies and improved the global consistency and local detail accuracy of the pseudo point clouds. The method proposed in the study provides a new approach for intelligent detection and recognition of power transmission lines, and provides a positive impetus for the fields of power engineering and computer vision.</p></div> |
|---|