Sub-meter Accuracy 3D Indoor Positioning Algorithm by Matching Feature Points of 2D Smartphone Photo

Abstract

Indoor positioning is a key technology for mobile advertisement and indoor Location-Based Services (LBS) that can rely on coarse localization accuracy of the order of ten meters. However, there is no satisfactory indoor positioning solution that can support high-accuracy prospective mobile services such as Augmented Reality (AR). In this paper, we developed a 3D indoor positioning method for smart phone users. This method is composed of two steps. In the first, offline phase, one uses an RGB-D camera to build a 3D indoor map, which is a collection of 3D feature points (FPs) of an indoor structure. In the second, online phase, a user device (such as a typical smart phone or tablet PC equipped with camera, 3-axis accelerometers and mobile data network) is geo-located. We designed an algorithm that calculates 3D position and attitude angles of the user device that captures a 2D photo and 3-axis accelerometer measurements. Experimental results showed that the positioning error of this method can achieve a planar RMS error below 10 cm. This accuracy is enough to support precise 3D Augmented Reality (AR) applications. In addition, this indoor positioning method can be easily implemented as a simple mobile application for smart phones or tablet PCs. In other words, this method does not need dedicated hardware or installing additional equipment. If Wi-Fi Access Points (AP) were used to obtain the rough position of the user device, and if cloud computing were used to reduce the computation time, the coverage of indoor positioning could be extended to large scale complexes such as shopping malls and downtown area buildings.