In the early summer of 2011 Terrasolid Ltd. performed a LiDAR survey of the City of Helsinki Tram Network. Helsinki Tram Company had three objectives for the project. Firstly, they wanted to get 5 cm accurate track geometry and other asset information to populate the maintenance management system theyre now piloting. The second objective was to get accurate curbs and the stops so that they could simulate how the 40 newly purchased trams would turn as they have a slightly different wheelbase, forward and rear overhang and ground clearance. The third objective was to find out how much high asphalt there is next to the rails as they had a project to change to wider wheels on the trams to improve the durability and lessen then noise in the turns. They wanted to know how much grinding work would be needed for different wheel widths. The final deliverable of the project consisted of the calibrated and classified point cloud and certain objects as a MicroStation DGN file in their coordinate system.
For Terrasolid this was a R&D project. We wanted to see how mobile mapping would perform in difficult city conditions where you have often very poor GPS satellite visibility. Also, a lot of new tools and features were developed in TerraScan, TerraMatch and TerraPhoto during the data processing.
The data collection
The data collection was carried out by 3D Laser Mapping (http://www.3dlasermapping.com/) from the UK. A StreetMapper was mounted on a tram. It consisted of two Riegl VQ-250 scanners, both at 200 kHz, a forward looking camera with 2144 x 1424 resolution and an IGI GmbH (http://www.igi.eu/) IMU/GPS.
The installation took place on Sunday the 29th of May and a test run was driven at the end of the day. The data collection started at 4 am on Monday the 30th but we had to return to base after 7 am when it started raining as the rain drops would have resulted in bad images. We continued at 4 am the next day and had almost the entire network scanned by 10 am but there were some loose bits here and there which took another two hours to reach due to the trams limited capability to reverse or turn around – we had to follow the network.
The total scanning time was approximately 11 hours. The driving speed was 25 km/h and due to the network structure some of the places were scanned numerous times. In total 14.6 billion points were collected; 0.5 TB in .LAS files and almost 20,000 forward looking images were taken – 15 GB in .JPG files.
The data processing
The data needed to be divided into blocks for processing. The high asphalt process could be run before tying the data into the coordinates as it was a relative process. A new tool was developed to automatically detect the rails from the point cloud. One first classifies the candidate points, in this case by using the trajectory as a guide and then defines the section template of the rails and which points are to be vectorized. Once run the system creates in this case three line strings at user definable steps. We are close to a release of a new tool which would then convert the centerline to design geometry ie. lines, curves and chlotoids. The high asphalt process produced a result that approximately 33.5 kilometers would need to be adjusted.
The next step was the data calibration. From previous experience we knew that there were many areas where satellite visibility was poor and where trajectory would rely on IMU only. Furthermore, we didnt have a DMI in the system. To conquer this we had targets painted in between the tracks fairly densely. In difficult areas there were signals every 50 meters and in open areas every 250 meters. A tool was developed to locate the paint markings automatically based on the echo intensity. Once the control measurements were received a fluctuating correction was applied to the point cloud. For Z-correction we used low altitude aerial LiDAR which is available for the whole city.
The initial estimated accuracy is shown in the picture below. Where the trajectory is painted red the estimated accuracy was worse than 15 centimeters, in open areas the blue trajectories had below 3 centimeter accuracy. During the data calibration we found that the trajectory had drifter 1.7 meters off, thus the high amount of targets proved to be invaluable.
Once the data was properly calibrated the best drive passes were chosen and the overlap was cut off. Then the automatic rail detection was run again now producing the rail geometry in the right coordinate system. For ground detection a new tool was developed to detect the dominant ground and a new tool was also developed to vectorize the overhanging wires. The curb stones currently have to be semi manually vectorized. Finally the point color can be applied to the points.
Conclusions
The project proved that mobile mapping can be effectively used to survey tram lines in difficult city conditions provided that there is sufficient targeting. The accuracy demand defines how dense the targeting needs to be.
In hindsight we could have used a StreetMapper on a car as well and would have been able to do the data collection faster having not been tied to the rails all the time. The initial thinking was that wed be able to derive the rail geometry from the trajectory but this proved to not work as the system was mounted in front of the tram and in the steep turns, a 10 meter radius is not rare, the forward overhang meant that the IMU path didnt follow the rails. That approach would work perfectly on a train survey.
About Helsinki Tram
The Helsinki City Transports Tram network (http://www.hel.fi/hki/hkl/en/HKL+Tram) is 97 km in total length. The tram is the main means of public transport in the inner city. Around 200,000 passengers use Helsinkis extensive network of tram lines every weekday. In 2008 a total of 53.9 million journeys were made by tram.
HKL Tram Transport runs the trams on behalf of HSL Helsinki Region Transport and is also responsible for maintaining and replacing rolling stock. They also perform a wide range of track and electricity infrastructure maintenance and repair services for HKLs Building Unit. In addition they are responsible for manufacturing track parts and for keeping the tram track clean and clear.
About Terrasolid
Terrasolid Ltd. (http://www.terrasolid.fi/en) is the developer of the world leading software suite for LiDAR processing. Its TerraScan, TerraMatch, TerraPhoto and TerraModeler are used in over 90 countries with over 4000 TerraScan licenses sold.
The author
Terrasolids business development manager Mika Salolahti, M.Sc. has over 25 years of experience in the field of CAD. He has worked for Intergraph, Mentor Graphics and 14 years in Bentley Systems before joining Terrasolid in 2009.