Study On Target Tracking Algorithm For Mobile Robot Based On Visual Attention Mechanism

Visual navigation is becoming increasingly important in the mobile robot navigation technology, The task of machine vision is to enable the machine observe objects as human, using machine vision to guide the robot to complete the originally need human beings to complete the work. A great number of studies have indicated that human have the ability, which is called visual attention mechanism that can guarantee the efficient work of human visual system, to select excellent visual data. The visual attention mechanism which can help to solve the problem of whether the robot can quickly identify the target from a large number of image data is introduced to the visual navigation of mobile robot. From the view of bionics, this paper use target tracking algorithm combined with human visual attention mechanism to explore the application of visual tracking of mobile robot.First of all,through the analysis of the characteristic of the mobile robot vision tracking technology and tracking system, combined with the requirement of vision target tracking system of mobile robot, it is concluded the advantage that human visual attention mechanism is applied to mobile robot target tracking system. At the same time through the test analysis of the current mature visual tracking method using problem in mobile robot, in order to analyze the feasibility of human visual attention mechanism is applied to mobile robot target tracking system. Then, based on human visual attention mechanism theory, comparative analysis of spectral residual method, frequency-tuned salient region detection, dynamic visual attention--searching for coding length increments and gbvs saliency map model of the four visual attention computational model. According to the requirement of the vision tracking system of the mobile robot, the model of selective attention computation model, GBVS saliency map, which is suitable for the visual tracking of mobile robot, is found.In order to carry out the experiment of mobile robot with selective attention mechanism of visual tracking algorithm, the visual tracking system of mobile robot is designed. First we set up a mobile robot platform according to the mechanical and electrical characteristics of the mobile robot design host computer image processing and control software, by designing the friendly human-computer interaction interface and user interaction, the entire syatem is buid. Then according to the GBVS saliency map model, This paper designs two kinds of visual object recognition and tracking algorithms:Target tracking algorithm based on selective attention model and mean shift algorithm based on visual attention mechanism, And the method of tracking algorithm integrated into the robot platform is proposed. Design a variety of actual conditions according to the actual working condition of the robot, each of these two algorithms were tested in the various conditions, through the test to verify the two algorithms of the accuracy, stability and real-time, get the characteristics of these two algorithms, and then propose the tracking algorithm, in order to improve the visual tracking system of mobile robot.This paper puts forward several method which simulate human visual attention mechanism for mobile robot visual tracking, and provides a way of thinking and algorithm framework for this kind of research. Finally we develope a visual tracking prototype system of mobile robot.