Airborne Lidar Power Line Classification Based on Spatial Topological Structure Characteristics

Abstract

Commission VI, WG VI/4KEY WORDS: airborne lidar, urban power line, neighbourhood selection, spatial topological feature, structure characteristics, power line classification ABSTRACT:Automatic extraction of power lines has become a topic of great importance in airborne LiDAR data processing for transmission line management. In this paper, we present a new, fully automated and versatile framework that consists of four steps: (i) power line candidate point filtering, (ii) neighbourhood selection, (iii) feature extraction based on spatial topology, and (iv) SVM classification. In a detailed evaluation involving seven neighbourhood definitions, 26 geometric features and two datasets, we demonstrated that the use of multi-scale neighbourhoods for individual 3D points significantly improved the power line classification. Additionally, we showed that the spatial topological features may even further improve the results while reducing data processing time.

Automatic Detection and Characterisation of Power Lines and Their Surroundings Using Lidar Data

“…The results of the classification are analysed attending to three parameters: correctness, completeness and quality, defined as shown in Equation 12: Table 3 shows that an average correctness of 99.24% was achieved, while the average of completeness and quality is 94.50% and 93.84%, respectively. The correctness achieved is higher than those reported by different authors, such as (Zhu , Hyyppä, 2014) (93.26%), (Chen et al, 2018) (96.50%), (Guo et al, 2016) (89.00%) and comparable to those found in (McLaughlin, 2006) (99.80%) and (Wang et al, 2017) (98.44%). The average completeness given in this paper is similar to those found in (Chen et al, 2018) (94.80%), and superior to (McLaughlin, 2006) (86.9%), (Guo et al, 2016) (86.00%) and (Wang et al, 2017) (83.08%).…”

“…A different method for power line extraction is shown in (Wang et al, 2017). A DTM is built to select candidate points.…”

Intelligent acceptance check for towers of overhead transmission line based on point clouds

“…This kind of methods need large calculations and are sensitive to interference such as trees or signal poles. Except for traditional feature-based methods above, supervised methods such as random forest [27] and support vector machine (SVM) [28] have also been tried to identify the tower. Deep learning also has been applied for classification of simple scenes [29][30][31], but the network training requires extremely high computing resources, if the training data are lacked, the generalization of the trained model cannot be guaranteed.…”

A LiDAR-Based Method to Identify Vegetation Encroachment in Power Networks with UAVs

“…On the other hand, in [18] ground and trees were identified based on calculation of geometric features using a variety of different neighborhood types and scales, suggesting that improved classification performance can be obtained with multi-type and multi-scale approaches, compared to singletype and single-scale alternatives. Along similar lines, in [19] Support Vector Machines were employed and a variety of geometric features were calculated considering multi-scale neighborhood schemes; thus, achieving precision rates of 95.64% and 93.83%, with the recall rates being 90.92% and 89.09% in two point clouds, respectively.…”