Abstract:In order to achieve the position estimation and tracking of targets on a ground plane,a theory for modeling the relation between the camera plane view and position estimates on a ground plane is proposed in this paper.Firstly,the transformation mode of random variables(position estimation on camera image plane and position estimation on a ground plane)under a projective transformation is analyzed,which shows that when some geometric properties are satisfied,a projective transformation maps normal distributions to normal distributions.Secondly,the unscented transformation is used to calculate the moments of the transformed random variables.Finally,a minimum variance estimator of the position estimation by multiple cameras is designed using the obtained modeling correlation,and is applied to the tracking of multiple targets in a dynamic environment on a ground plane.The experimental results show that the proposed model not only has good combined position estimation ability,but also can effectively present and track targets on a ground plane by using the minimum variance estimator obtained by this model.