Research On Indoor SLAM Algorithm Based On Binocular And Inertial Measurement Unit Fusion

As the population ages and labor costs increase,indoor mobile service robots and robots that provide convenient services in public places are becoming more widely used.In fact,the current level of intelligence of robots cannot meet people's needs.Among them,sensing the surrounding environment and making judgments is one of the key technologies to achieve intelligence.Therefore,in recent years,simultaneous positioning and mapping of robots(Simultaneous Localization and Mapping,SLAM)has been a research hotspot.Although with the improvement of computing capabilities of mobile devices and the development of computer vision theory,robot localization and mapping technology have made great progress,but they are still susceptible to the surrounding environment and cost constraints.At present,robots cannot complete tasks independently and accurately.Therefore,in this subject,the binocular camera and IMU sensor information are fused to a mobile robot to improve the robot's autonomous perception of unfamiliar environments.During the loop detection of the robot,the bag of words model was redesigned and trained to improve the response speed of the robot system in the indoor environment.The specific research contents are as follows:The imaging and projection estimation models of binocular vision SLAM system are established to provide a basis for subsequent vision and IMU information fusion and calibration experiments.By studying how to extract the ORB feature points in the image and how to optimize the matching results of the feature point pairs,the visual SLAM system can detect whether it has passed through the same place,thereby eliminating the cumulative error during the movement.Based on the ORB_SLAM2 model,build a VI-ORB-MBag(Visual-IMU ORBSLAM Modified-Bag)model.This framework realizes the fusion of binocular visual information and IMU information,and solves the deficiency that the monocular SLAM system must be initialized to obtain the distance information of three-dimensional points.The VI-ORBMBag model uses the complementarity of the vision sensor and IMU information to improve the robustness of the system.In order to improve the system's speed of loading bags of words,ensure real-time operation on mobile devices,and re-collect small-scale image data.Compress and preprocess the collected images,and set the bag-of-words tree structure parameters and evaluation functions.Then retrain the bag-of-words model to improve the robot's perception of the surrounding environment.Finally,a laptop and a binocular camera with an IMU sensor are used as experimental platforms.Test the loading speed and response time of small-scale bags of words under the same conditions.Test and compare the robustness of the ORB_SLAM2 and VI-ORB-MBag systems and the errors of the sparsely constructed trajectories under standard data sets and real environments,respectively.