The RGB-D SLAM Combined With Monocular Vision And Its Implementation Under Cloud Robot

The research of Simultaneous Localization and Mapping(SLAM)is of great significance for robot to achieve fully autonomous movement.In recent years,with the development of sensor technology,the acquisition of the depth image has been more convenient,and it makes more and more people pay attention to RGB-D SLAM.Based on the present research status of SLAM technology,this paper compares the different Monocular SLAM algorithms in visual SLAM,and analyze the RGB-D SLAM algorithm.Based on these theories,we proposed an enhanced RGB-D SLAM combined with monocular vision,and we also extends the algorithm,so that it can be used in Cloud Robot environment.The main innovations of this paper are as follows:(1)、This paper presents a method of RGB-D SLAM,which aims at solving the problem that depth camera can’t obtain the depth information and the depth information can not fully acquired due to the rapid movement of the camera.The proposed method combined different advantages of monocular vision SLAM and RGB-D SLAM.Firstly,extract the feature points of the RGB image,and count the Feature points combined with Depth image.In this way,it can judge whether the depth information is acquired or not.Secondly,through a selection control unit,when the depth information can’t be obtained or the depth information is not completely obtained,the monocular vision SLAM method will be used to build the Monocular Local Map;when the depth information is sufficient,RGB-D SLAM method can be used to construct RGB-D Local Map directly.Finally,to construct the Global Optimal Map by concatenating and merging the Monocular Local Map into RGB-D local map.This method makes all the depth camera has a wider application space,not only can it be used in the narrow indoor environment,but also can be extended to spacious indoor scene.At the same time,this method can improve the stability of RGB-D SLAM method,and build the Global Optimal Map more quickly.(2)、On the basis of above research,by applying the cloud computing technology,it allocated the complicated and expensive map optimization process,together with calculation tasks to a service in the Cloud,while the high real-time tracking tasks are assigned to run on local robots.During the process of working,the robot client sends the key frame to the cloud server through a standard wireless network and completes the tracking task with the local sub-optimal Map;while the cloud server constructs the global map after receiving the key frame,and also optimizes and fuse the local map to the global map,then sends the final map to the robot client.The combination of cloud computing and RGB-D SLAM can not only improve the efficiency of SLAM,but also reduce the robot load and size,so that robots can be used in more complex environment.