Research On The Monocular ORB-SLAM Positioning System With Real Scale

In the current indoor engineering construction scene,in order to improve the safety and efficiency of workers,the unmanned and automatic engineering vehicle is of great significance,especially in some occasions where pollutants need to be treated,the operators should be avoided to enter the construction site as far as possible.In order to realize automatic operation,the positioning system is the key.A positioning system with high precision and strong stability is an important guarantee to improve the working efficiency of the vehicle.This paper presents a monocular ORB-SLAM positioning system with real scale information to realize the positioning function of engineering vehicle.The specific research contents are as follows:Taking the small crawler excavator on the market as the prototype of the operation vehicle,the automation transformation was carried out,and the signal converter,wireless communication module and photoelectric switch and other sensors were installed,so that the upper computer can remotely control the output of the proportional valve of the operation vehicle,get rid of the original manual or handle remote control operation mode,and realize the remote unmanned operation of the operation vehicle.The vision SLAM scheme is adopted in the positioning system of unmanned operation vehicle.The monocular ORB-SLAM algorithm is the core algorithm,and only the monocular camera is used as the only sensor of the positioning system.In order to improve the positioning accuracy and robustness of the whole system,and solve the scale uncertainty of traditional monocular vision SLAM,an absolute depth measurement method based on image segmentation is proposed in this paper to calculate the real depth of some objects in the image in static scene.The depth measurement method is integrated into the initialization stage of monocular ORB-SLAM,and the measured absolute depth value is used to modify the point cloud depth value after triangulation of the feature points of the initial image frame,so that the SLAM system can get the real scale.Experimental results show that the improved ORB-SLAM positioning system can effectively reduce the absolute trajectory error,and can obtain the trajectory map with real scale information in real-time operation.The positioning accuracy of the whole positioning system is close to that of stereo SLAM,and can solve the scale uncertainty of monocular vision to a certain extent.