Mobile Robot Localization Based On Visual Tracking

Localization of the mobile robot is a hot research topic now, this article adopts the method of visual tracking and a global visual model for mobile robot localization, which results in the real-time spacial coordinates of the robot. Experiments show that the visual tracking and localization system has a good real-time performance and high accuracy.Due to the fixed size of the search box in the Meanshift algorithm, Camshift algorithm is a better choice. For the shortcoming of missing tracking when the Camshift algorithm is applied in occlusion cases, this paper studies the prediction mechanism of Kalman filtering, and finally proposes the improved Kalman filtering-based Camshift tracking algorithm. Under the condition of occlusion, this algorithm can predict the current state model on the basis of historical measurements, then combine the predicted values of the previous frame with the observations of the current frame according to the weight coefficient to calibrate status model.Real-time visual tracking carried out through the camera usually gets an image coordinates, then how to convert the image coordinates to spatial coordinates is another priority we need to study. First, we use OpenCV vision library in VS2010 development platform for the camera calibration. Second, we establish a global visual model based on the planar motion of the mobile robot, which is to set the Z plane as zero. Thus a directly inversed perspective mapping from the image coordinate system to the world coordinate system can be build. So long as the inversed perspective matrix calculated, we can achieve real-time physical coordinates of the robot.Finally, we do some experiments on the robot combined visual tracking method with the global visual calibration model described in the above, and achieve good performances in the real-time visual localization process.