An application of data fusion combining laser scanner and vision in real-time driving environment recognition system

Abstract

This paper describes a newly developed combined calibration method for laser scanner and monocular vision in real-time driving environment recognition system. First, a special right angled isosceles triangle calibration board was designed to calibrate the extrinsic parameters of laser scanner including mounting height and pitch, in this way the laser points can be converted into the vehicle coordinate system. Then, the characteristic points in laser data and image data can be extracted to build a constraint equation to obtain the parameters of the pinhole camera model. Finally, the laser points can be projected onto the image with this model and displayed in the same image frame. Experimental tests on running vehicle showed that the system could avoid interference, ambiguity and display the driving environment including obstacles, vehicles, pedestrians and so on in real-time. The calibration process is very simple and comprehensible, and the calibration board is quite easy to obtain. With less computation load, our work can expedite the obstacle detection and classification.