The Research On The Localization For Indoor Mobile Robot Based On Binocular Vision

With the rapid development of science and technology,the smart mobile robot has made a figure all walks of life.The robot technology is developing and improving continually from the simple mechanical structure only in repeating motion for people at the beginning to the smart mobile robots which possess autonomous perception and making decision now.It also has a great effect widely on social production and human activity.The localization plays a great important role in executing motion by robots and is a prerequisite for robots making next decision.The binocular vision possesses advances of a wide range of perception and large quantities of information compared with traditional equipment such as compass,inertial navigation system,wheel odometry and so on.Binocular vision also becomes a significant research field and immensely improves the accuracy of localization for robots.The mobile robot is mainly divided into indoor robot and outdoor robot based on the environment.This paper is aimed at the indoor circumstance and embraces the technologies of embedded system,binocular vision system,binocular visual odometry and so on to develop the platform of software and hardware.This paper also makes a study on localization for indoor mobile robot based on binocular vision.This paper designs an indoor mobile tracked robot upon the embedded system according to the demand.It contains installation of operation kernel,setting of wireless network,the model selection and access of perceptive sensors for environment,setting of control kernel and control motor.This paper realizes the robot embedded software system,reading the information of sensor,transportation of control command and interactive interface upon web based on already set up hardware system.Such interactive interface upon web obtains robot of the information and controls robot of the mobility in time.Then the research on localization is made in the own set up mobile robot system.First of all,the theory of binocular vision system is elaborated including the calibration of camera and The principles of perception of tridimensional environment by binocular vision.After that,the feature point extraction and matching algorithm are analyzed.The SURF algorithm is selected as method of feature point extraction by comparing the experimental results.Then the algorithm of binocular vision odometry based on feature point matching is payed more attention and discusses how to realize the localization by means of feature point extraction and matching,feature point tracking and motion estimation.The binocular vision odometry localization algorithm is also accomplished at own set up platform of mobile robot.Finally,this paper show some group of experimental results to verify the feasibility of robot system and algorithm by comparing the results of localization for wheel odometry.