With the development of science and technology,mobile robot has become the more and more popular research object all over the world.There are many key issues in the research of the field of mobile robot,but the accurate localization of the mobile robot is the premise for the mobile robot to complete a set of tasks.There are many different methods of mobile robot localization.Different methods of localization of mobile robot have different characteristics.Considering the existing laboratory conditions,comprehensively comparing of several mobile robot positioning methods,we decided to research the localization of mobile robot based on visual tracking.A three-wheel omni-directional mobile robot is used for this topic experiment platform.The Mobile robot motion controller is NI myRIO,and each wheel motor is equipped with a photoelectric encoder.There are a red mark and a blue mark above the mobile robot,and they are respectively used for mobile robot visual tracking and calculating angular displacement of the mobile robot.We can use a computer to get image form the camera,which is fixed on the wall,and we can observe the real-time information of the state of the mobile robot.In the experiment,Camshift visual tracking algorithm is used for the localization of mobile robot research.But because that there are always some obstructions,which can cause shade to track the target,in the visual tracking process,and it can make visual tracking fail.To solve the problem,the algorithm of Kalman with a feature of forecast mechanism is introduced to combine with the algorithm of Camshift to research the localization of mobile robot.Experiment shows that using Camshift tracking algorithm based on Kalman filter is a good way to solve the problem of obstructions.According to the result of the analysis of the experiment data,the localization result of the mobile robot based on visual tracking is very close to the real value,whether displacement on the X and Y direction,or angular displacement.The localization error based on visual tracking method is much less than the method based on odometer.The localization error based on visual tracking of displacement on the X and Y direction has remained in the range of 20 mm,and the angular displacement error keep within 0.2 rad and odometer(encoder)positioning error will accumulate with the change of the time.It has achieved a good effect to use the method of visual tracking to research the localization of mobile robot. |