Research On Monocular Vision Trajectory Tracking Method Based On Extended Kalman Filter

Trajectory tracking refers to the problem of positioning and trajectory description of autonomous mobile robot in motion.The robot has a very broad application prospect,which is widely used in many disciplines.It combines the advanced research results in various fields,and reflects the strength of the scientific research in various countries.In these researches,the application of ultrasonic sensor,encoder,laser sensor,vision sensor and other sensors are included.Among these sensors,the monocular vision sensor has many advantages such as large amount of information,cheap price and strong error tolerance.So it has a wide range of research on saving the cost and improving the positioning efficiency.In today's research,the traditional single sensor is difficult to meet the requirements of positioning accuracy,multi-sensor data fusion technology began in the academic world.In this paper,a kind of SLAM(Simultaneous Localization And Mapping)alogorithm is proposed.It is based on monocular vision and inertial measurement unit.The algorithm mainly solves the problem of positioning robot and building map in the environment of GPS(Global Positioning System)signal loss.Firstly,we improve the traditional vision position with the monocular visual odometry technology.Secondly,we introduce the inertial measurement unit as auxiliary equipment to predict position of the robot so that it can improve the accuracy of robot localization,solve the problem of drifting error of acceleration measurement and the image loss in visual SLAM.Thirdly,using the EKF(Extended Kalman Filter)model to fuse the position of the robot and 3D acceleration information,and try to get more accurate location information to increase the robustness of the algorithm.Finally,to verify the effectiveness of the localization and mapping process,we build the experimental platform of visual and IMU positioning system,and the experiment is based on the measured data.The result show that the proposed algorithm is superior to the traditional monocular vision localization algorithm in robustness and accuracy.