Research On Localization And Mapping Of Indoor Mobile Robot Based On Vision

Mobile robots' perception of unknown environments is a key step towards intelligence,and it is also a research hotspot and difficulty in the field of robotics.Simultaneously Localization and Mapping technology can effectively solve the key problem of mobile robots' environmental awareness.With the development of computer vision and sensor technology,visual SLAM has received widespread attention.The current visual SLAM scheme still has the problem of poor accuracy and susceptibility to interference from dynamic objects,which affects the system.Therefore,the vision-based indoor mobile robot positioning and mapping research in this paper has important research significance and practical application value.The main research contents of this article are as follows:In order to solve the problem of missing part of depth image,which is caused by noise and limitation of Kinect camera.A repair strategy based on Gaussian filtering and multi-frame geometric information method is proposed.This strategy preprocesses the original depth image through Gaussian filtering,and repairs the missing part of the depth image by combining the depth information of the same feature point between adjacent frames.Aiming at the dynamic interference sources dominated by indoor mobile pedestrians,this paper proposes a visual SLAM system for dynamic environments based on the feature method and the YOLO target recognition algorithm.It realizes indoor pedestrian target recognition of a single RGB image by training a deep convolutional network model.The feature points within the range of the recognition frame are removed to ensure the accuracy of pose estimation,which eliminating the interference of indoor mobile pedestrians on camera positioning.The experimental results under the public datasets show that the SLAM system proposed in this paper has centimeter-level positioning accuracy in the presence of indoor pedestrian interference,which is significantly better than the standard ORB-SLAM2 system and has more accurate positioning accuracy.Combining the ROS system and Turtlebot mobile robot to complete positioning and mapping tasks in real indoor scenes.The experimental results show that the vision-based SLAM system proposed in this paper has good dynamic environment adaptability,achieving real-time performance.