Road Markings Segmentation from LIDAR Point Clouds using Reflectivity Information

TL;DR

This work addresses robust road-marking segmentation from LiDAR data by exploiting the reflectivity channel of a 64-layer sensor rather than raw intensity. It introduces a pipeline with pre-filtering, road-plane segmentation, region-growing clustering, per-layer adaptive thresholding on reflectivity using Otsu, and iterative line fitting, implemented in C++ with PCL. Experiments on a test track and on highways show reflectivity-based segmentation outperforms intensity-based processing across lighting conditions, achieving high precision and F1 scores. The approach enables reliable multi-lane road-marking extraction for lane-change planning, with future work on tracking over time, curved-road modeling, and evaluation under adverse weather.

Abstract

Lane detection algorithms are crucial for the development of autonomous vehicles technologies. The more extended approach is to use cameras as sensors. However, LIDAR sensors can cope with weather and light conditions that cameras can not. In this paper, we introduce a method to extract road markings from the reflectivity data of a 64-layers LIDAR sensor. First, a plane segmentation method along with region grow clustering was used to extract the road plane. Then we applied an adaptive thresholding based on Otsu s method and finally, we fitted line models to filter out the remaining outliers. The algorithm was tested on a test track at 60km/h and a highway at 100km/h. Results showed the algorithm was reliable and precise. There was a clear improvement when using reflectivity data in comparison to the use of the raw intensity data both of them provided by the LIDAR sensor.

Figures (6)

• Figure 1: The same point cloud from the test track was colored based on the level of reflectivity (a) and level of intensity (b).
• Figure 2: Flow graph of the developed algorithm.
• Figure 3: Point cloud output at the different stages of the developed algorithm. (a) In red is the output of the pre-filtering and plane segmentation stage. The output (red) of the region growing algorithm is depicted in (b). In (c) we can see the road-markings candidates (red) after the adaptive thresholding. Finally, in (d) we can see the final road marking (red points) with the associated lines (red lines) as well as the ones that were rejected by the algorithm (light blue).
• Figure 4: The JKU-ITS research vehicle Certad2022 was used to collect the data.
• Figure 5: (a) and (b) depict the point clouds (dark gray) recorded when traversing the 6-lane segment of the test track. Red points are the detected road markings and blue lines indicate the line models. (c) and (d) shows the onboard camera images at the same time that (a) and (b) were recorded. In (c) the sun was in front of the vehicle making lane determination difficult with an image-only approach.