Research On Elevator Maintenance Assistance System Based On Markerless Augmented Reality

With the innovation of hardware and sensor technology in the 21 st century,the augmented reality technology to get rid of the bondage of heavy equipment,as a display device is applied in many fields,and become one of the supporting technology of contemporary industrial 4.0,greatly reduce the burden of the cognitive load and knowledge workers,however,with the constant extension augmented reality applications,Traditional augmented reality technology has been difficult to cope with diversified application environments and requirements,and augmented reality is prone to tracking failure and virtual model registration problems in extreme environments,which reduces its most important appearance and practicability as a display method.Therefore,how to improve the adaptability of augmented reality to specific application environment and the stability of model registration is the key research direction to ensure the stable display of augmented reality.This paper studies the problem of unmarked augmented reality elevator maintenance assistance system,and the main work is as follows:Firstly,for the visual SLAM module of unmarked augmented reality,the original feature extraction method has the problems of less feature points extraction and poor matching effect in the uncertain environment of elevator shaft illumination.In feature extraction stage,before feature point extraction,the image frames of elevator shaft environment obtained by mobile camera are processed with gray scale and adaptive histogram equalization accelerated by interpolation is used to improve image quality,make feature point extraction more stable when the brightness changes and reduce the overlap and aggregation of feature points.In the feature matching stage,the RANSAC algorithm with low iteration times and polar geometry constraints is used to eliminate mismatches.It can be seen from the test that the method is better than the original algorithm in extracting and matching elevator shaft environment features,which lays a foundation for the subsequent calculation of augmented reality pose and point cloud data.Secondly,in terms of information interaction between the visual SLAM module of augmented reality and the rendering engine.In order to solve the problem of model slosh caused by the change of plane parameters during the frame-by-frame plane fitting of point cloud data in unmarked augmented reality,a point cloud splicing module based on point cloud overlap rate was proposed to process the continuous frame point cloud and realize smooth transition of continuous frame fitting plane.Aiming at the problem that the plane fitting accuracy of point cloud is poor due to the influence of outliers and rough margins in point cloud data,the adaptive threshold RANSAC algorithm is used to eliminate outliers in point cloud data,and PCA method is used to perform the second plane fitting,so as to improve the fitting accuracy of placing surface of virtual model.Experiments show that the proposed point cloud splicing module performs well in real-time and accuracy,and the plane fitting accuracy is better than the traditional algorithm.Finally,the mobile terminal augmented reality elevator maintenance auxiliary system is implemented.The elevator maintenance support system based on improved the logos of augmented reality architecture,visual SLAM module used for elevator well environment improvement of feature extraction and matching algorithms,augmented reality module USES improved anti-noise plane fitting method,from the visual SLAM module,point cloud splicing module,augmented reality realization of system function module and man-machine interactive module,And demonstrate the system.