Map-based Real Time Indoor Navigation Using Smartphone

Abstract

This paper presents a method for building a map-based real-time indoor navigation system using a smartphone. Generally, the GPS sensor is used to estimate the location, but since the GPS sensor error occurs inside the building, the user’s initial position and real-time movement are determined using the smartphone internal IMU sensor and camera instead of GPS. The neural network learns images taken with a camera and detects objects in photos taken by users based on the learned information. The initial position of the user is obtained by matching the image of the detected object in the picture with the image of the corresponding label. Then, the user’s posture is estimated using the IMU sensor, and the low pass filter is used to analyze the pattern of the filtered data to predict the walking interval when the smartphone is held in the hand. After that, the movement path of the pedestrian was calculated using the filtered acceleration and the angular velocity of each step. Finally, the location correction was performed by projecting the calculated route according to the user’s actual position estimation on the map data including the movable route information. The optimal route of the destination is searched using the Dijkstra algorithm.