A combined motion and appearance model for human tracking in multiple cameras environment

Inter-camera color calibration for object re-identification and tracking

“…In previous work [1], we introduced a 1-D color based appearance model which preserves objects spatio-color information. This feature allows recognizing an object that exits and re-enters in the field of view of different cameras.…”

“…Object position prediction helps to estimate future objects position in the video frames and also can give some objectobject occlusion prediction [1]. We use Kalman filter [12] to calculate the probabilistic position and to detect occlusion of objects in next frame.…”

“…In this paper, we investigate object identification and tracking in multiple non overlapping cameras separated by blind regions. We have previously proposed [1] an appearance based object tracking algorithm having the ability to reidentify the moving objects. Our tracking process is applied independently to each single camera, but the objects features are stored in a common object database which allows object recognition when the object exits from the field of view of one camera and then enters in another cameras.…”

Multi-Camera Tracking Helmet System

“…Conversely, Lian et al [2] try to tracking objects in multiple non-overlapping cameras, they proposes a Bayesian model to solve the consistent labeling problem across multiple non-overlapping camera views. In [3], a combined motion (position and velocity of object) and appearance model for human is built in database to help recognize objects in different camera views. Work in [4] presents a video summarization method to visualize the video object trajectories across multiple cameras in a static image for monitoring the  Manuscript received February 10, 2013; revised April 16, 2013. movements of suspicious people in a building, human feature based method is used to associate objects.…”