In November, OxTS’ regional support team in China received an invitation from the Site Operations department of the Sichuan Test Base for Sino-German Intelligent Connected Vehicles (hereinafter referred to as the Sino-German Test Site) to establish pointcloud data in the core test area of the Sino-German Test Site.
The reason for the survey was to support pre-research work in the areas of scene library establishment, automatic cloud testing, digital twin testing, and digital proving ground management.
The Sino-German Test Site is located in Chengdu, Sichuan. It is the 8th national test demonstration base project approved by the Ministry of Industry and Information Technology.
It covers an area of about 1,305 acres and has 9 major functional areas, including:
- A 500 m underpass tunnel
- A 3.6 km closed ring highway
- A large curvature interchange
- Two 200 m, three-lane rain and fog environment simulation tunnels
For the data collection work, the Sino-German test site put forward the following technical requirements:
- Pointcloud data must include all test-related elements in the test area, including traffic lights, signage, road markings, etc.
- The pointcloud accuracy of non-tunnel scenes needs to be better than 10 cm.
- Each point in the pointcloud data needs to have positioning information, and the coordinates are defined as a plane (XY) coordinate system. The RT in the future test site will be able to match the pointcloud data after inputting the same coordinate origin and direction information to achieve matching between the test vehicle position and the pointcloud image.
- Pointcloud data must be in a common format to facilitate analysis and application by third-party software.
- The pointcloud data file should be as small as possible and can be loaded and viewed normally on an ordinary laptop.
In order to meet the requirements set by the Sino-German test site, we used OxTS’ flagship product the RT3000 v3 and Hesai Technology’s XT32 lidar for data collection.
Accurate Calibration
The first step in pointcloud data collection is the pose calibration of inertial navigation and LiDAR. Accurate pose calibration can not only perform motion compensation and remove image distortion, but also ensure the splicing accuracy of multi-frame pointclouds. OxTS’ unique boresight technology uses a data-driven approach to automatically calculate and complete the alignment of inertial navigation and LiDAR.
The following figure shows the difference in scanning the same area before and after boresight calibration…
After precise calibration, the pointclouds collected by LiDAR from different viewing angles can completely overlap without ghosting or deformation.
Read the OxTS Boresight Calibration brochure to learn more about the process – Boresight Calibration Brochure
Accuracy Guaranteed
The Sino-German test site has a wealth of tunnel and elevated scenes. In order to ensure the accuracy of the pointclouds in these scenes, we used OxTS LiDAR Inertial Odometry (LIO) technology on some road sections. This technology uses post-processing to obtain velocity data using the difference between LiDAR frames, and then provides it to the inertial navigation engine as auxiliary information. In the case of poor satellite signals, position drift can be greatly suppressed and positioning accuracy improved. Compared with wheel speedometers, LIO technology allows the inertial navigation engine to obtain speed information with six degrees of freedom (3 linear speeds, 3 angular speeds), and more precisely controls the calculations in each direction.
Read the OxTS LiDAR Inertial Odometry technical article to learn more about the feature – OxTS LIO Technical Article
Pointcloud mapping and voxelisation processing
Using OxTS Georeferencer, pointcloud data can be processed after importing inertial navigation data and LiDAR data. The software supports the output of common file formats such as LAS and LAZ, making it easy to import third-party software for other applications. In order to reduce the file size, OxTS Georeferencer supports point cloud voxelisation processing, automatically retaining the most accurate points within a spatial grid of defined size, eliminating other points, and obtaining a smaller pointcloud file without losing accuracy.
Download the OxTS Georeferencer datasheet for more information – OxTS Georeferencer Datasheet
With the strong support of the Sino-German test site, OxTS successfully completed the collection task and delivered all pointcloud data of the core test area of the Sino-German test site in late November.
Scroll through the pointcloud screenshots below and see the final results…
Jeremy Li
OxTS Regional Support Engineer.
