Indoor visible light positioning system based on image sensor
Date of Issue2018-09-10
School of Electrical and Electronic Engineering
In this dissertation, a visible light-based IPS using commonly used LED lamps and smartphone camera is investigated and implemented in Android phone. An ID extracting algorithm called frequency-label based on 2ASK-modulation  is designed to facilitate the extraction of ID frames from captured pictures by leveraging the rolling shutter effect of CMOS camera. In the procedure of position estimation by solving nonlinear equation sets, an optimized algorithm based on non-iterative SVD-based method  is deployed to obtain the close-form least square solution to position matrix and rotation matrix. Also, optimization of an image sensor-based indoor visible light positioning (VLP) system  by improving the positioning algorithm is discussed. We implement the algorithms on an Android phone by JAVA programming with the basic idea of state machine to control functional modules (capturing, image-processing and calculation). Real-time experiments are carried out to verify the performance of the proposed indoor positioning system. Results show that 3-D positioning errors are 6 cm on average, in the experiment space of 2×2×2 m3. The results reveal that the introduced SVD-based noniterative algorithm is more time-saving than the conventional algorithm (approximately 50-80 times). Meanwhile the positioning error and performance of the optimized VLP system are investigated experimentally. This dissertation achieves smartphone-based real-time indoor 3D positioning of high robustness with centimeter-level accuracy and 2 Hz positioning rate.
DRNTU::Engineering::Electrical and electronic engineering