Calibration of Inverse Perspective Mapping from Different Road Surface Images

Abstract

Inverse perspective mapping (IPM) is a geometrical image processing technique to remove the perspective distortion in an image and transform the image as if seen from the bird's eye view (BEV). IPM can be used as one of the key features in the advanced driver assistance system (ADAS) in vehicles. IPM requires camera calibration in order to extract intrinsic and extrinsic parameters of the camera. However, IPM assumes flat road surfaces. The accuracy of the IPM decreases if these parameters are constant when dealing with variety of road surfaces. This paper presents an implementation of camera calibration technique with incorporation of selected point features from different images containing different road surfaces. Results show that the generated IPM images calibrated with different road surfaces, such as straight, hilly, and arterials (bumpy and zebra crossing) roads, produces enhanced visual appearance, less blurry and less pixelated.