Research And Inplementation Of Visual Object Tracking Method For Mobile Robot

It is feasible to operate independently intelligent robot with the development of semiconductor technology,the improvement of sensor performance and that of memory and CPU.In 2011,the Fukushima Nuclear Power Plant in Japan leaked,resulting in significant losses,because no massive rescuing measures could be taken by human rescuers in person,which made it out of control.From now on,Countries began to pay much more attention to developing robots which can replacing human's work in harsh environments.In 2012,The Ministry of Science and Technology of the People's Republic of China issued a document which point out that it is necessary to focus on the development of intelligent control technology for service robots.Robot Object Tracking is a technology of robot intelligent control,it has wide application prospect in the field of service robots.Since service robots mainly services for humans,this paper study pedestrian tracking for robots.In the existing methods,the robot object tracking has been achieved by using sonar sensors,laser radar,binocular camera and RGB-D camera.These methods have their own shortcomings and can't meet the requirements of practical application,so that this paper will propose a method which can meet the practical application requirements and using monocular camera to achieve the robot object tracking.In this paper,the existing DSST and TLD monocular vision object tracking algorithm has been tested on the robot.It is found that the DSST algorithm has no tracking failure detection mechanism and can't deal with the problem of complete occlusion of objects.On the contrary,the TLD algorithm has the tracking failure detection mechanism and can deal with complete occlusion of objects,but it can't tracks pedestrian well in the real scene.The changing light condition,the object fast motion and the object non-rigid deformation are likely to lead to a tracking failure of TLD algorithm.After theoretical analysis and experimental results comparing,it is found that the advantages and disadvantages of DSST and TLD algorithm are complementary,so that this paper combined these two algorithms into a new algorithm named dsstld.Experimental results shows that the dsstld algorithm has better tracking performance than DSST and TLD algorithm.In addition,this paper further improves the detection module of dsstld algorithm,then proposed a better algorithm named dsstldh.This paper not only research the improvement of object tracking algorithm used in robot object tracking,but also implement a complete and easy-to-use mobile robot object tracking system.The system integrates object tracking,speech recognition,face detection,map building and autonomous navigation technologies,which can control the robot to switch its states in static,tracking and navigation by voice command,effectively improved the human-computer interaction convenience and the intelligent level of the robot.