Indoor 3D Localization Using Monocular Vision

At present, global positioning system (GPS) basically meet the user's needs for location services in outdoor scenes. However, it is difficult to locate where GPS signals are missing in indoor environment. In this paper, we design an indoor localization system based on monocular vision. Firstly, according to the pictures taken by the terminal camera in real time, the feature points of each picture are extracted, and then the position and pose of the monocular camera are estimated by using the direct method to match the features. In order to estimate the three-dimensional depth of feature points, this paper use the depth filter to establish a real time sparse point cloud map. Finally, the real environment is simulated by using the three-dimensional visualization tool RVIZ. The experimental results show that the positioning accuracy of the proposed method in indoor environment can reach 0.04m.

I. Introduction

In recent years, Indoor localization technology has been developed rapidly [1] [2]. However, it is difficult to locate in indoor environment where GPS signal is missing. On the other hand, traditional indoor positioning technology mainly uses Bluetooth or WIFI and other methods to integrate into a set of positioning system. However, this method is highly affected by the environment, and often requires deployment of multiple devices. In addition, this traditional positioning method can only get the distance and angle information, but the pose of the device in 3D space cannot be obtained.