Summary: | 碩士 === 國立臺灣大學 === 資訊網路與多媒體研究所 === 103 === This thesis presents an augmented reality indoor navigation system with invisible polarizer marker and visible light communications (VLC). Polarizer is a polymer multi-layer composite materials. It has a physical property that
human eyes cannot distinguished the intensity of polarized light passed though by only one polarizer. But the intensity can be discriminated and controlled by adding another polarization filter with a different polarization direction.
VLC is a new wireless communication technology. LED has gradually replaced the traditional fluorescent lighting because its advantages of long life, low power consumption, and high efficiency. Due to its properties of high
energy efficiency and fast response time, LED has a great potential for indoor positioning.
In our proposed system, we use commodity camera as the receiver and modify off-the-shelf LED light to become the transmitter, and thus it requires minimal additional cost. The main idea is that when the camera captures images
with the polarizer marker on the ceiling light fixture, the orientation angle between the camera and polarizer marker can be estimated. Utilizing VLC, at the time the LED also transmits its absolute location to the camera. The device
can then use these two pieces of information to estimate its own absolute location when more than one LED light is captured in the image.
This solution is highly attractive for any indoor environment with LED lighting, due to our design goal is to be able to estimate the orientation of the
device with a single image capture. In this thesis, we present two methods to estimate the orientation angle between marker and the camera using a single image: the ratios between the pixel intensity of different marker areas, and the orientation of the marker in the image. We also propose a method to mitigate
the error with multiple images. Our system is evaluated with different image resolutions and distances between the light and the camera. Results show that the mean angle error is always less than 10◦ for a distance up to 2.5 meters.
|