Research On AGV Autonomous Localization Technology Based On Lidar And Vision Fusion

Today,intelligent manufacturing technology is developing rapidly,and at the same time,the demand of manufacturing enterprises for intelligent transportation business continues to grow.In this environment,researchers have developed a new type of intelligent transportation equipment,namely Automated Guided Vehicle(AGV),which can complete transportation work autonomously and flexibly,and significantly improve the intelligence level of logistics operations.In complex work scenarios,AGVs can discover flexible and low-cost logistics paths through Simultaneous-Localization-and-Mapping(SLAM)technology,which is extremely important to the development of the logistics industry.Based on this,this paper mainly studies how AGV can perform autonomous positioning in unfamiliar environments,discusses the practicability of SLAM technology,and what are the shortcomings of SLAM technology,analyzes the problem of AGV relocation at runtime,and proposes specific solutions.This research mainly includes the following contents:(1)Analyze the perception principle of 2D lidar sensor.For out-of-plane obstacles,it is difficult for 2D lidar to perceive and scan effectively.In this regard,combined with multi-sensor information fusion technology,visual information is integrated to improve the perception effect of 2D lidar.Through fusion data enrichment The visual information can make the mapping more complete and reliable.(2)Analyze the traditional laser SLAM algorithm framework.In the work of 2D lidar,if the amount of information obtained is insufficient,it is difficult to quickly judge the closed-loop point,which makes the mapping deformed.In this regard,this paper improves the closed-loop detection method,and combines visual information to judge the closed-loop point.,through the specific experimental verification and analysis,it is found that this method can significantly improve the deformation problem of the map and make the map more accurate.At the same time,in order to make the positioning of SLAM technology more accurate,the author proposes a joint optimization method,which increases the constraints by fusing a variety of information,thereby improving the positioning accuracy.Through specific experiments,this method can make the positioning of the laser SLAM technology more accurate.Compared with the traditional algorithm,the positioning accuracy can be maintained within 4cm.(3)Analyze traditional visual retargeting methods.Since the relocation method is extremely dependent on the AGV path,it is difficult to achieve global relocation.In this regard,this paper proposes a corresponding improvement strategy,using the bag-of-words(Bo W)model combined with the traditional positioning technology.The global map range is relocated.Through specific verification experiments,compared with the traditional visual positioning technology,the recall rate of this method is improved by 26.1%,and the accuracy rate of translation error is maintained in the range of 2cm-4cm,and the range of rotation error is 1.5°-2.5°.