Vehicle association among multiple cameras

In this paper, we propose an active method to associate the behaviors of the target among multiple cameras. We place a sample collection of photographed vehicles in each camera, which is captured by its own. The captured target is used to calculate a set of distance vectors with the sample collection in each camera. To improve the accuracy of distance vectors matching between images from different viewpoints, we propose a method combined color histogram and ASIFT to obtain the distance vectors. Afterwards we compute the distance of the two vectors to determine whether the targets in two cameras are the same. In no overlapping region case, there may be some mistakes among similar vehicles-the latter cameras may associate with another object which has the exact same features with the target. In this article, we use a prediction method, which utilizes spatial and temporal properties to solve this problem.