Visualization of 3D cable between utility poles obtained from laser scanning point clouds: a case study

摘要

We can automate inspection work of infrastructure facilities by analyzing the characteristics of 3D structure information obtained through 3D structure visualization using a point cloud. The safety level of equipment can then be diagnosed quantitatively. In this paper, we investigate the modeling of wire structures such as overhead communication cables between utility poles, which are close to the ground, have many obstructions, and have a complex structure. We evaluate the accuracy of cable models and compare them to the correct model. We use three modeling methods: a machine-learning method based on the extruded surface of a point cloud as a feature, a rule-based method involving principal component analysis, and models generated from a combination of these models. In addition, we focus on modeling overhead cables from field data (urban and suburban). Results show the practicability of modeling overhead cables with a cable length of 10-70 m regardless of the area type. We find that the best cable modeling rate with the precision and recall of 80.76% and 83.84%, respectively, can be obtained using the machine-learning method and by specifying the cable reproduction rate to be 2 m.