3D AVM system for automotive applications

The car body structure and vision limitations of the rear view mirror may lead to inner wheel difference and blind spots, which results in traffic accidents. To solve this problem, many automakers develop the Around View Monitoring (AVM) actively. The AVM system provides the information around the car for drivers with videos. This paper is based on the 2D AMV and further develops the 3D AVM, which uses four fisheye cameras to set in the front, in the rear, and in the rear view mirrors of both sides to record videos. This paper applies fisheye correction, perspective transform, image stitching, and scale-invariant feature transform to synthesize four input videos into 2D AVM, and the 3D AVM is created by 2D AVM to 3D AVM algorithm. Compared to the 2D AVM in previous methods, the 3D AVM proposed in this paper can help drivers to determine the collision distance while crossing with another car in narrow roads and parking, which can prevent traffic accidents. Moreover, drivers can switch perspectives arbitrarily according to the requirements without the limitation of fixed overlooking view. Therefore, the driving safety is improved.