There is an increased utilization of the mobile terrestrial laser scanning (MTLS) systems in different road corridor applications. These systems are fast and accurate; allowing very high density point clouds to be acquired. But their use is still limited due to their cost and the huge amount of data they capture. Processing that huge amount of data is extremely labor intensive, time consuming and requires a lot of manual processing.
The aim of this research are to automatically detect different road furniture such as
poles, curb and the street floor. This will be done by developing new methods for the
automated segmentation of those features from a 3D point cloud captured by a MTLS system.
Automating the analysis of the data reduces human bias, and both the field scanning
and the point cloud processing can be conducted more rapidly. The MTLS systems are
expensive, but normally the cost of utilizing them and the point cloud processing software
is less expensive than the equivalent analysis using traditional methods like utilizing the Total Station and GPS.
Automatically identifying the road poles from the MTLS point cloud is very important
and will make the detection of the attached objects easier. The road curb and the street
floor represent very important road furniture. The curb separates the street floor and side walk; it is also used to direct rainwater into the drainage system. The automatic detection of curb points from the MTLS point clouds helps in defining the road boundaries and the curb condition.
Automatically detecting a highly detailed street floor helps in maintaining the pavement by estimating the road surface conditions. The location of bumps and dips can be detected to estimate the roughness of the road surface.
In this research different point cloud processing pipelines have been proposed and
implemented. These pipelines have successively detected the road poles, the road curb and the street floor from different MTLS point cloud scenes. The detection is done just based on the input 3D point cloud and without utilizing any additional data source.
Four different point cloud datasets have been tested with the proposed methodologies.
These datasets were captured by the TITAN MTLS system. The datasets represent
different road scenes, like downtown and highway areas. They have different point cloud densities and side objects. These varieties enable testing the performance of the proposed methodologies.
The results show the efficiency of the different point cloud processing pipelines and its applicability with different road scenes. Also, the proposed curb and street floor detection pipelines do not require any additional information about the road, like its geometry and direction. This additional information need other data sources such as the trajectory of the scanning system, aerial image or a map.
The main contribution of this research is to automate most of the processing steps
of the point cloud captured by MTLS systems. Automating the detection of different
road furniture like the poles, the curb and the street floor is important step in order to get the full benefits of the MTLS systems. This is done through developing a new point
cloud processing pipelines which are applicable to be utilized on different scenes, and automatically setting most of the thresholds for the utilized detection parameters.