Classification of 3D Point Cloud Data from Mobile Mapping System for Detecting Road Surfaces and Potholes using Convolution Neural Networks

Normally, road damages can be automatically detected using image and video data from ground survey vehicle system combined with the detection algorithms. However, there are limitations of scales map coordinates when to detect potholes. It has been challenging determine sizes locations This research utilized a mobile mapping system, MMS, collect roads environment classify potholes, other objects. A convolution neural network (CNN) was used directly identify 3D point clouds XYZ method in comparison proposed XYZ-RGB method. The classification demonstrated an overall accuracy 96.77%, intersection over union (IoU) roads, objects 59.50%, 94.22%, 94.06%, respectively. indicated 97.50%, IoU 66.66%, 95.43%, 95.42%, Both datasets were statistically compared at 95% confidence level, results revealed that both classifications produced significantly different results.