Parking assistance based on time-to-flight cameras

The main objective of the parking assistance systems is to help the driver park, verifying that there are no objects at the rear of the vehicle and hence avoid potential accidents. Most of these applications are based on ultrasonic sensors, wide-angle lens cameras, RADAR or LiDAR.

At TECNALIA, we have decided to go one step further, designing a parking assistant based on the reconstruction of objects around the vehicle, using information obtained from a Time-to-Flight (ToF) camera.

This technology detects registration plates of nearby vehicles using the relationship between pixel and 3D point to estimate the position, predict collisions and detect free parking spaces.

This system has already been tested in parallel parking scenarios, such as bay parking, obtaining a striking 93% precision. In this way, this technology has proven increased reliability in the detection of registration plates thanks to the dual verification (image processing and point cloud), as it minimises false positives. Furthermore, it improves the detection of surrounding vehicles and, as it uses the distance between consecutive vehicles, it can also identify free parking spaces.

On the other hand, a 3D reconstruction of the scenario provides better assistance than conventional methods, thanks to its path estimation and 3D collision warnings that lighten the burden on the driver in parking manoeuvres.

The use of a single sensor for image and depth processing minimises the development and implementation costs.

We will present the results of this research at the Automation Seminars that will be held from 4th to 6th September in Ferrol. This event is a meeting point for automation researchers, teaching staff and professionals.