CLASSIFICATION FOR ROADSIDE OBJECTS BASED ON SIMULATED LASER SCANNING
Kenta Fukano1 and Hiroshi Masuda2
1) Graduate student, Department of Intelligence Mechanical Engineering, The University of Electro-Communications, Tokyo, Japan.
2) Prof, Department of Intelligence Mechanical Engineering, The University of Electro-Communications, Tokyo, Japan.
Abstract: The maintenance of roadside objects, such as utility poles, traffic signs and streetlights, is very important for sustaining infrastructures, but it is costly and tedious work to survey a huge number of roadside objects. A mobile mapping system (MMS) is promising for improving the efficiency of maintenance tasks. In our previous work, we proposed a machine learning approach for automatically identifying roadside objects from point-clouds captured by a MMS. However, supervised machine-learning methods require training data, which have to be carefully created by manually classifying roadside features. In this research, we propose a method for automatically generating training data using 3D CAD models. Point-clouds of roadside objects are created from CAD models by simulating laser scanning on a moving vehicle. In our experiments, our method could generate reasonable point-clouds that can be used as training data for supervised machine learning.
Keywords: point-cloud, mobile mapping system, machine learning, classification
Kenta Fukano, and Hiroshi Masuda. “CLASSIFICATION FOR ROADSIDE OBJECTS BASED ON SIMULATED LASER SCANNING.” In Proceedings of International Conference on Civil and Building Engineering Informatics (ICCBEI 2015), 110. Tokyo, Japan, 2015.