Research On Laser And Visual SLAM Fusion Algorithm Based On Improved EKF

At present,the warehousing industry widely uses automated guided vehicles(AGVs)for cargo handling,so under the condition of reducing human participation,improving the efficiency of AGVs in the warehousing environment has received great attention.In this process,the positioning,mapping and automatic identification and positioning of the AGV and the insertion pallet are the key steps to achieve the efficient and accurate work of the unmanned AGV.At present,the positioning and mapping of AGVs mainly use single-modal laser Simultaneous Localization and Mapping(SLAM)and visual SLAM.Among them,singlemodal laser SLAM technology can effectively extract the depth information of feature points,and visual SLAM can use rich texture information;However,laser SLAM performs poorly under unclear environmental characteristics,dynamic environmental conditions,and poor repositioning ability,which makes it difficult to restore the working state after tracking is lost.Visual SLAM performs poorly in environments with no texture or weak lighting,and monocular cameras have the disadvantages of requiring initialization,uncertain scale,and scale drift when the distance is unknown.With the development needs and trends of larger scale,higher flexibility and higher efficiency of domestic logistics warehousing,the integration of laser SLAM and visual SLAM has become a hot research topic.Therefore,this paper uses the multineo-information theory to improve the Extended Kalman Filter(EKF)fusion laser and visual SLAM data,firstly,the multi-neo-information theory can effectively use the data of historical moments,improve the accuracy of the corner point data after fusion,and then realize the detection of tray holes according to the color image.However,AGV throwing in the storage environment has the following challenges,so this paper proposes different solutions:(1)Aiming at the problem that the existing EKF fusion laser and visual SLAM data methods only use the current moment data,resulting in low accuracy of corner point estimation,this paper proposes an improved EKF fusion laser and visual SLAM data method based on multiple new information.The multi-neoplasmic theory can effectively use the data of historical moments to improve the accuracy of corner point estimation fusion of laser and visual SLAM in each frame.The experimental results show that the average errors of the proposed method are 0.0268 and 0.0109 at 20 iterations and 100 iterations,respectively,in the indoor environment,and the accuracy of corner point estimation is improved by 33.9% compared with the unimproved EKF method.(2)Aiming at the uncertainty problems such as uneven light and complex background in the storage environment,this paper proposes a tray hole detection method based on color image.The method is divided into two stages: camera and AGV relative position calibration and tray hole recognition.The purpose of the relative position calibration of the camera and the AGV is to solve the problem of the correct transformation of the pallet pose to the AGV coordinate system under the camera coordinate system.Tray hole recognition mainly uses normal filtering to reject noisy point cloud data,crop out the effective area of the pallet,use the threedimensional point cloud to realize the coarse detection of the tray hole,and then identify the pallet outline through the color image,match the target contour and the multi-scale template profile to obtain the accurate pallet area in the scene,and verify the effectiveness and rationality of the method by experiments.(3)Combined with the above-mentioned improved EKF fusion laser and visual SLAM data algorithm and pallet detection and pose calculation algorithm based on multiple new information,this paper designs an unattended AGV automatic cargo handling system based on Intel NUC industrial computer and QT framework.In the warehousing environment of the listed enterprise Dongjie Intelligence,the effectiveness and reliability of the system have been verified.In this paper,the method of multi-information improvement EKF fusion laser and visual SLAM data and pallet detection method are studied,and an unattended AGV automatic cargo handling system is designed,which verifies the effectiveness and reliability of the proposed method and system through experiments in the actual storage environment.