Research On Mapping And Navigation Algorithms Of Mobile Robot Based On Vision

Robot technology integrates mechanical control,sensor technology,signal processing,bionics and other disciplines.With the rapid development of cloud computing,artificial intelligence,big data and other technologies,robot technology has become one of the most active fields in the high-tech frontier technology.It is not only used in the traditional industrial manufacturing field,but also in the fields of entertainment,transportation,military defense and medical health.As one of the important branches of robot technology,mobile robot integrates environment perception,path planning and behavior control,which makes mobile robot technology gradually become the biggest research hotspot in robot field.Mobile robot technology mainly needs to solve three problems: the location of mobile robot,the acquisition of surrounding environment information and path planning.An improved orb-slam algorithm is proposed to solve the problems existing in the existing simultaneous localization and mapping(SLAM)methods,such as large matching error,slow running speed and low positioning accuracy.Firstly,a full convolution neural network is used to preprocess the scene image and extract the significant region in the image.Secondly,the image pyramid method is used to construct the scale space,and fast corner detection is carried out in the scale space to obtain the location and scale information of the image feature points.Thirdly,in view of the large number of error loops in the original ORB-SLAM algorithm,a loopback detection algorithm based on the scene image region of interest and hierarchical weighted matching ideas is proposed to improve the accuracy and recall rate of loopback detection.The experimental results of visual odometer based on open data set show that the matching accuracy is improved by 37% and the efficiency is improved by 44% compared with the results of feature extraction and matching for the whole image.The back-end experiment results show that the proposed algorithm has better accuracy and real-time performance than the original ORB-SLAM and two improved algorithms based on ORB-SLAM.In this thesis,an improved A~* algorithm is proposed to solve the problems of low operating efficiency,not smooth generating path and poor robustness of A~* algorithm in path planning algorithm.By optimizing the storage structure of the OPEN table,the time of path planning can be effectively reduced without increasing the number of access nodes.According to the experimental simulation results,the path planning time of the improved A~* algorithm is about10% less than that of the original A~* algorithm.Aiming at the problem that the mobile robot keeps wandering in front of obstacles,A~* algorithm is combined with the local path planning algorithm,and the effectiveness of the algorithm is verified by the actual scene experiment.