A classification model for power corridors based on the improved PointNet++ network

Li Bo,Liu Siyuan,Wang Fengxiang,Zou Cunyu

DOI: https://doi.org/10.1080/10106049.2023.2297556

IF: 3.45

2024-01-19

Geocarto International

Abstract:Aiming at the existing deep learning classification model for power corridor point cloud still need to improve the classification efficiency and the robustness of the classification model to meet the requirements of practical applications. An improved classification model based on PointNet++ is proposed. Based on the fact that the main features of the power corridor scene are power lines, poles, and vegetation, the initial data are first optimally filtered, and then the ensemble abstraction module of the classical PointNet++ is modified to better adapt to the power corridor scene. Finally, h-Swish is used as the activation function to realize the accurate classification of the features of the power corridor scene, and the training time of deep learning is also greatly reduced. The experimental results show that the improved algorithm achieves an average F1 value of 97.58%, which is 3.62 percentage points higher than the classical PointNet++. Therefore, the algorithm has great potential in point cloud classification.

geosciences, multidisciplinary,environmental sciences,remote sensing,imaging science & photographic technology

What problem does this paper attempt to address?

### Problems the Paper Attempts to Solve

This paper aims to address the issues of low classification efficiency and insufficient model robustness in existing deep learning classification models for power corridor point cloud classification. Specifically, the authors propose an improved PointNet++-based classification model to meet the needs of practical applications.

### Summary of the Main Content of the Paper

1. **Research Background**:
- The application of UAV inspection technology in power transmission line detection is becoming increasingly widespread, enabling comprehensive monitoring and rapid analysis, thereby improving inspection efficiency and safety.
- High-precision 3D point cloud models are one of the key components of refined autonomous UAV inspections.

2. **Current Status and Challenges**:
- Currently, many scholars have conducted relevant research on the classification of power transmission line point cloud data and have proposed various methods, such as PointNet++, SparseCNN, KPConv, etc.
- The existing PointNet++ model has some limitations when dealing with power corridor scenarios, especially in terms of local feature extraction.

3. **Improvement Methods**:
- An improved PointNet++ model is proposed, which first performs optimal filtering on the initial data, and then modifies the set abstraction module in the classic PointNet++ to better adapt to power corridor scenarios.
- Using h-Swish as the activation function, the model achieves precise classification of power corridor scene features and significantly reduces the training time of deep learning.

4. **Experimental Results**:
- The experimental results show that the improved algorithm achieves an average F1 score of 97.58%, which is 3.62 percentage points higher than the classic PointNet++.
- The improved model demonstrates higher accuracy and stronger robustness in point cloud classification tasks.

Through the above improvements, the model has great application potential in power corridor point cloud classification tasks.