University of Missouri simplifies lidar assessment of road traffic

01 Apr 2025

Single lidar system and AI analysis could remove need for multiple light sources.

A project at the University of Missouri has investigated whether a combination of lidar and AI could address key issues in transportation safety and mobility.

Lidar for automotive monitoring and control is a major focus of research, although recent geopolitical tensions and tariff uncertainties have led to challenges in the market.

In March 2025 two of the lidar industry's key players, Luminar and Ouster, announced major changes to their product portfolios in recognition of the volatile business climate.

Missouri's project was intended to address some of the inherent complexities and constraints involved in using lidar for traffic volume data collection, as a route to less costly systems. The findings were published in Journal of Transportation Engineering.

"The emergence of modern technologies, particularly lidar, has revolutionized traffic monitoring by enabling efficient and accurate data collection," commented the project in its paper. However, despite the benefits of using lidar for traffic data collection, "previous studies have identified two major limitations that have impeded its widespread adoption."

The first hurdle is the need for multiple lidar systems to obtain complete point cloud information of objects of interest, by triangulating different perspectives on the scenes of interest. The second is the labor-intensive process of annotating three-dimensional (3D) bounding boxes, as part of object detection.

The Missouri team developed an innovative framework that both alleviates the need for multiple lidar systems and simplifies the laborious 3D annotation process. This approach should reduce the data acquisition cost and address its accompanying limitation of missing point cloud information, by developing a point cloud completion (PCC) framework to fill in any missing data.

Better understanding of how pedestrians and cyclists interact

In trials, the project set up a joint camera and lidar system at an intersection to monitor traffic flow. Although only one lidar system was viewing the scene, the team developed a framework for extracting features from the object of interest, such as height, acceleration and speed.

Combining the extracted height information with a 2D bounding box derived from the image allowed 3D bounding boxes to be generated automatically, without human intervention.

"Instead of retraining a machine learning model to detect objects, we used a pre-trained one and created a new algorithm to estimate an object's height and width," said Yaw Adu-Gyamfi from Missouri's College of Engineering. "This helped us classify objects, such as buses, pedestrians and cyclists, more accurately than other AI models designed for the same task."

Next steps in the research will involve considerations of power supply stability and weather conditions, but ultimately the data collected by this system could then be used for other purposes, such as tracking cars entering work zones, catching speeding or distracted drivers, or spotting pavement problems such as the depth of potholes.

"By having a better understanding of how pedestrians and cyclists interact with each other on the roads, this study will help us design advanced systems that will allow vehicles to better understand and avoid other road users," said Yaw Adu-Gyamfi. "This is important especially as autonomous vehicles become more common."