LiDAR gives an accurate picture of surrounding environment, opening new possibilities

lidarPoint CloudUnikie AI Vision

Share this story: 

Share on facebook
Share on twitter
Share on linkedin

LiDAR gives an accurate picture of surrounding environment. This is good for many tasks which have previously required human eye. Unlike human stereo vision, LiDAR uses lasers and time of flight based measurements to determine the distance to target object, or visible surface of that object – to be exact. This opens new possibilities for many kinds of human assistance and automation applications. Good example is automatic obstacle detection and warning system for vehicles.

The basic idea is easy, let the LiDAR work as eyes of the vehicle. LiDAR gives a 3D snapshot of the surrounding environment, called a point cloud. This is not enough, however, since the system is still missing a critical part, the brain. In this analogy the brain would be the algorithms responsible for processing the stream of point clouds to ultimately determine if there is an incoming collision.

A real obstacle detection done by LiDAR is much better than a simple parking sensor found in modern cars. Regular parking sensors work reliably only with short distances, as they simply tell if there is something in front of the sensor up to a given distance. This kind of approach would not be suitable for longer distances, while for instance, if a perfectly good road becomes steeper it would sound the alarm, since the sensor sees the road itself as an obstacle, see Figure 1. Similarly, when a downhill begins, the sensors point up in the air and would miss any real obstacles on the road.

Figure 1: Estimating the ground level from the LiDAR point cloud.

Using a LiDAR allows for much more precision, but also requires much more complex algorithms for detection than just a proximity sensor. Our Unikie algorithms detect the road regardless of its relative orientation to LiDAR. After that it detects any obstacles that would block or be harmful for the vehicle and keeps track of them, see Figure 2.

Figure 2: Detecting an obstacle on track with estimated ground level.

If such an obstacle is on collision course with the vehicle, a suitable feedback is provided. The feedback can be anything from an audible warning identification to enforcing full autonomous braking of the vehicle. The tracking of objects over multiple measurements also greatly reduces the number of false positives from noise in such a system.

– The basic idea is easy, let the LiDAR work as eyes of the vehicle. This technology opens up new possibilities for many kinds of assistance and automation applications, says Tuomas Pöyhtäri, Unikie LiDAR algorithm specialist.

More Stories

Data Science, Machine Learning, Predictive Maintenance
Business, Data Science, Machine Learning, Predictive Maintenance


This website uses cookies to ensure you get the best experience on our website. By clicking “Got it” or by continuing to browse our website you agree to the terms we have set out for collecting user data and storing it in our database(s). Please see our Privacy Policy here.