The Key Technology For Autonomous Localization Based On Vision And IMU

The problem of autonomous localization of ground robots and UAV is the basis of navigation and path planning,and it is also a hot research direction in the field of artificial intelligence.The method of relying on external device localization in an unknown environment may be limited.Therefore,autonomous localization based on the sensor carried by itself is a prerequisite for the robot and the UAV to perform tasks in a complex environment.Therefore,the main research contents of this thesis are vision-based robot autonomous localization and visual-IMU-based autonomous localization of UAV.The key technologies of vision-based autonomous localization and autonomous localization based on visual and IMU are studied.The specific research content of this paper is divided into the following parts:(1)Firstly,the background and significance of autonomous localization of moving agents are investigated and analyzed.Then the application of SLAM is researched.It is clear that the SLAM method is necessary.(2)The autonomous localization platform for ground robots and UAV are designed and built for different autonomous localization methods.Considering the application scenarios,space size,carrier load,size and computing power,the ground robot autonomous localization platform and the autonomous localization platform of the UAV are designed for different application scenarios.(3)The key techniques of autonomous localization method based on vision are researched?Then an improved autonomous localization method is proposed.The front of system is improved,the image processing method is added to remove the influence of the environment to the picture,and an improved feature matching method is proposed.Then the experiment was carried out on the indoor robot autonomous positioning platform.Firstly,the improved method was tested to prove that the image quality increased and the matching correct rate increased.Then the improved method and the original method were located based on the indoor positioning system.The comparison of accuracy proves the improvement of precision.Next,a number of experiments were carried out on the robot to realize the autonomous positioning of the robot.(4)Aiming at the exiting problem of vision-based autonomous localizationmethod,the visual-inertial odometry is proposed,then the method is verified in the UAV's autonomous localization platform.The experimental results show that the proposed algorithm has a certain improvement from the visual autonomous positioning method in terms of robustness,stability and positioning accuracy.And it can better meet the needs of the autonomous localization of the drone.Finally,the autonomous positioning of the drone is realized based on the autonomous positioning method whether in the indoor environment or the outdoor environment.