SLAM Technology Of Regional Intelligent Electric Vehicle Based On Multi-sensor Fusion

In recent years,the technology of artificial intelligence has developed rapidly,and it has received extensive attention and applications in the field of mobile robots and intelligent vehicles.Simultaneous Localization and Mapping technology,referred to as SLAM technology,has widely attracted researchers' attention as an important technology in these fields.This paper studies the limitations of the laser SLAM.By incorporating visual information into the laser SLAM solution,its processing capability in loop closure detection is enhanced,thereby improving the accuracy of the SLAM system loop closure detection and building a more accurate global consistent map.First of all,research is carried out on the LOAM method that performs well in laser SLAM,and the principle,framework and characteristics of the scheme are analyzed.Using open source laser point cloud data to make a data set in the Bag package format to test LOAM's point cloud registration,lidar odometer,lidar mapping and transformation processing module.The test results show that LOAM can complete the positioning and mapping in real time and accurately,but there are stills deficiencies in dealing with the closed loop scenario.Then,to solve the problem of LOAM's lack of laser loop closure detection module and the unsatisfactory accuracy in closed scenes,a laser loop closure detection module is designed for LOAM.By extracting the key frames in the laser data frame and comparing the similarity of the key frames to determine whether the loop closure detection occurs,the Gtsam optimization library is used to optimize the results of the loop closure detection.The experimental results show that after adding the loop detection module,the consistency of the constructed map is significantly improved compared to the map without adding the loop detection module;however,when dealing with scenes where the carrier moves too fast,the success rate and accuracy of loop closure detection remains to be improved.Finally,in order to further improve the accuracy and environmental adaptability of laser SLAM loop closure detection,a visual information assisted loop closure detection method is proposed.On the basis of extracting the ORB feature points of visual images,key frames are selected to generate ORB word bags.The matching of key frames and word bag models is used to complete the visual loop closure detection.The results of the visual loop closure detection are used to modify the results of the laser loop closure detection.The experimental results show that after the fusion of visual loop closure detection,the success rate and accuracy of the loop closure detection are improved,and it can cope with the scene where the speed of carrier movement becomes faster.