| Automated Guided Vehicle(AGV)is a kind of logistics transportation equipment and it has intelligent and automatic features.Indoor AGV has been widely used in modern intelligent logistics storage systems.Simultaneous Localization And Mapping(SLAM)is the key technology in AGV autonomous navigation.At present,the navigation methods of indoor AGV mainly include: electromagnetic navigation,inertial navigation,navigation and positioning based on twodimensional code,and visual navigation.Compared with other navigation methods,the navigation method based on vision is flexible,autonomous,intelligent,etc.With the improvement of computer vision and image processing technology,the visual-oriented navigation method has made great progress and it has been a hot spot and focus in the research of AGV navigation.There are three main types of cameras in visual navigation: monocular cameras,binocular cameras,RGB-D cameras,RGB-D is a kind of depth cameras,compared with monocular and binocular cameras,it can easily obtain the depth information corresponding to image pixels by means of Time of Flight(ToF).It is very convenient and suitable for positioning and mapping research.In the past,scholars used monocular and binocular cameras for indoor positioning and mapping.The RGB-D camera was applied to outdoor positioning and mapping research,but no navigation research for indoor AGV was found.This paper innovatively uses RGB-D camera to study the simultaneous localization and mapping of indoor AGV,so that AGV can automatically locate and construct the surrounding environment map in real time in an unknown environment,and improve its perception.The SLAM system mainly has two parts: the front end and the back end.The front end is a visual odometer.Firstly,it uses the image information to extract the feature points,and then uses the three-dimensional space geometry to estimate the motion of the camera,the back end includes three parts: nonlinear optimization,loop detection and mapping.Nonlinear optimization generally uses Gauss Newton,Levinberg-Marnelt and graph optimization to optimize the results from the visual odometer,loopback detection uses a dictionary model to decrease the error caused by the camera sensor and the target is to obtain a globally consistent result.The mapping is to construct the environmental information in the form of map.This paper systematically studies the SLAM system algorithm and design an RGB-D SLAM system,which is applied to the indoor AGV.The SLAM system is implemented by expected,and the AGV can autonomously synchronous positioning and mapping in unknown indoor environment.At the same time,the research of this paper provides an overall framework for the subsequent development and design of SLAM system based on RGB-D,and has made a certain practical basis for subsequent research. |
