Research On Object Tracking Algorithm For Mobile Robot Based On Visual Attention

For mobile robot, acquiring information by vision is one of the most important sensing methods. The ultimate intention of robot vision research is to replace people on work by robot, which can imitate the function of human vision system. Just as proved in the research, human have extraordinary ability in data filtration, and the vision attention is the basic mechanism to ensure this kind of high efficiency. The vision attention mechanism is also valuable in solving the problem of how to recognize the object faster from the large amount of image data in robot navigation by vision. The mainly purpose of this paper is to research the application of robot vision in object tracking based on human vision attention mechanism.The structure of the mobile robot vision system in the research is introduced at the first, and then an overall frame of the research is designed, according to the tracking task and the experiment condition. The system is divided into many basic blocks whose mainly functions and realized approaches are discussed.In object recognition block, an object attention model is designed based on the theory of vision attention, which integrates the image-driven, bottom-up attention and the object-driven, top-down attention. By the bottom-up attention process, the whole scene is segmented into the ground region and the non-ground region. The ground region is the area which robot can move, and the non-ground region consists of the salient regions. By the top-down attention process, the feature model of object is established. And then, the object region is gained from the salient regions by matching with the object feature model.In order to acquire the spatial position and motion parameters of moving object, the kinematical model of the vision platform and the vision projection model of the camera are both analyzed, and then the transformation matrix of the image pixel coordination and the world coordination is established. By the reverse projection model, the object's world coordination can be acquired from object's image pixel coordination. According to the status of robot, object tracking algorithm is then designed to control the mobile robot's movement.The experimental results show that the algorithm is effective in mobile robot's moving control. The accurate recognition and speedy tracking of moving object are realized.