Skip to Main Content
Vision based localization techniques in an intelligent environments have been well studied. But it still faces several difficulties, such as limited surveillance area, low response speed, relative high computation cost and so on. In this work, we create a large planar camera array which extremely extends the scope of surveillance area, and more precisely localizes robots with high response speed for AGV system. In this camera array system, color based tracking algorithm efficiently performs the task of localizing and identifying robots. A hybrid tracking algorithm intelligently controls the camera array and significantly reduces the redundancy of image data to be processed. Therefore, the whole camera array can be controlled by one normal PC rather than the expensive and powerful computer cluster or embedding the micro-processor into cameras. The system also can automatically re-calibrate cameras if the pose of camera changes during an accident. Experiments show the camera array consisting of 15 cameras can localize 5 robots with 20 Hz~60 Hz.