A multisensor fusion approach to shape estimation using a mobile platform with uncalibrated position

Abstract

Estimation of three dimensional shapes is a difficult problem which has been considered by researchers in machine vision and robotics. A typical problem is to estimate shape using images taken from various viewpoints. We consider a camera mounted on a mobile platform which is actively controlled effecting changes in the viewpoint of the camera. The mobile platform also has the capability of 2D ranging in addition to the camera-hence providing an opportunity for sensor fusion. In this paper, we first developed a new technique to calibrate the position of the mobile platform based on range lines and odometer. Then based on these calibration parameters, the shape (plane) estimation using extended Kalman filter (EKF) is carried out utilizing multiple views from the CCD camera and the range data from the range finder. The effectiveness of this approach has been proved by experimental results.