LiDAR for bike detection

Student thesis: Master thesis (including HD thesis)

  • Malthe Sigurd Birkemose Holm
4. term, Transport Engineering, Master (Master Programme)
This thesis examines the use of LiDAR for bike detection in traffic. Bike detection has a wide array of applications, including statistics of bike use, traffic adaptive signal control and other applications in ITS (Intelligent Transport Systems). The purpose of this project is to determine the capability of LiDAR in detecting cyclists travelling in close proximity in order to assess the number of cyclists, their positions and to calculate speed and ETA based on this.

Data is collected using a HDL-32e LiDAR mounted on a lamppost for detection of cyclists on a bike path leading up to a signalized intersection in Aalborg, Denmark. Collected data is evaluated using the Pylidartracker program which has been specifically designed for this thesis. Data analysis is done using in-house developed Python scripts.

During the timeframe of data collection, 233 cyclists were observed using a video camera. With the LiDAR-data, Pylidartracker and Python scripts, it was possible to determine the precise position and calculate speed and ETA, to a defined point, for all 233 observed cyclists. In one instance, two cyclists in very close proximity to each other led to an overestimation of the number of cyclists by 1, making the number of detected cyclists 234. The study therefore achieved an overall precision on bike detection with LiDAR of 99.5\%.

This study has some limitations, as it was only possible to achieve the reported accuracy within a 12 meter range of the LiDAR. Furthermore, the data collection was done during COVID-19 restrictions, which may have affected the number of cyclists as well as their proximity to each other.

This study concludes that LiDAR is highly capable of detecting individual cyclists within a range of 12 meters, but that further studies are needed to determine the capability of LiDAR in detecting cyclists travelling in close proximity in groups. This study includes proposals for designs of future studies to avoid the limitations faced in this study.
Publication date10 Jun 2020
Number of pages113
ID: 333939912