Vehicle Detection Using Probability Fusion Maps Generated by Multi-camera Systems

Abstract

In this paper we describe a multi-camera traffic monitoring system relying on the concept of probability fusion maps (PFM) to detect vehicles in a traffic scene. In the PFM, traffic images from multiple cameras are inverse perspective-mapped and registered onto a common reference frame, combining the multiple camera information to reduce the impact of occlusions. Although the unconstrained perspective projection is non-invertible, imposing the condition that the image points be co-planar allows the relaxation of this constraint. In images of road scenes, the road surface is locally planar, and as such can be inversely mapped. We show that computing the perspective undistortion by finding 4 matching points between a camera image and an ideal non-perspectively distorted image, a system of linear equations can be solved that corresponds to applying the rotation, translation, and re-projection onto the common reference plane without needing calibrated cameras. We show that this method yields good results in the detection of vehicles for subsequent tracking and monitoring.