AVM Auto-Calibration for driving vehicle images

Abstract

AVM is widely used in Advanced Driver Assistance Systems. The conventional AVM algorithm performs distortion correction, and transforms the corrected image to a bird’s eye view image using perspective transform. However, we have to manually obtain four coordinates from the image to complete the transforming matrix. In this paper, we propose an auto-calibration method which extracts the required four coordinates by using the number plate of a vehicle and the vanishing point of an image. For 640x480 fish-eye lens input images with a deviation rate of ±5 degrees, we could succeed in restoring the deviated image’s bird’s eye view.