Creating 360 View by Combining Multiple Fish-eye Images

Nielsen, Freddy Gramstrup

4. term

Vision, Graphics and Interactive Systems, Master

2016

2016-06-02

64

At kombinere billeder fra flere bilmonterede kameraer for at give en 360-graders visning i fugleperspektiv til hjælp ved parkering er en forholdsvis ny idé. Selvom det tidligere er blevet undersøgt, er det først i de senere år, at bilproducenter er begyndt at sælge sådanne systemer. Dette projekt undersøger brugen af fire fiskeøjekameraer til at skabe en komplet 360-graders visning. Hovedtilgangen er at kalibrere kameraerne for at bestemme linseforvrængning og orientering, rette billederne (korrigere forvrængning), lægge dem over på et fælles jordplan (vejbanen) og samle dem til ét billede. En sekundær tilgang blev også undersøgt, hvor inputbillederne i stedet projiceres direkte på halvkugleformede overflader i et virtuelt miljø for at skabe et samlet billede. Resultaterne fra hovedtilgangen viser, at det er muligt at lave en 360-graders visning med kun fire fiskeøjekameraer. Den sammensatte visning kan hjælpe føreren ved at afdække områder, som ellers ville være blinde vinkler, hvis man kun brugte ruder og spejle.

Combining images from several cameras mounted on a car to provide a 360-degree, bird’s-eye (top-down) view for parking assistance is a relatively new idea. Although it has been studied before, only in recent years have car makers begun selling such systems. This project investigates using four fisheye cameras to produce a full 360-degree view. The main approach is to calibrate the cameras to estimate lens distortion and orientation, rectify the images (correct distortion), align them to a common ground plane (the road surface), and composite them into a single image. A secondary approach was also explored, in which the input images are projected directly onto semi-spherical surfaces in a virtual environment to produce a composite. Results from the main approach show that it is possible to create a 360-degree view using only four fisheye cameras. The combined view is expected to help drivers by revealing areas that would be blind spots when relying only on windows and mirrors.

[This abstract was generated with the help of AI]

Camera Calibration ; Image Processing ; Image compositing ; vehicular safety

Download
View record in AAU Student Projects