Recently, with the advancement of technology, Intelligent Transportation Systems(ITS) and autonomous driving systems have attracted much attention. In this work, we studied on the Optical Camera Communication (OCC)-based V2I system to provide smooth traffic and road environments in autonomous driving applications.
We used a camera attached to the vehicles as the receiver of OCC communication structure and LED infrastructure on the road as the transmitter. In the transmission system, OOK (On-Off Keying) method was used, and a Manchester code was applied to maintain the light function of the infrastructure. The receiving method is to use the image processing technique for the input of real time or video file. The receiver also used image processing technique for detection and tracking to infrastructure light sources. ROI(Region Of Interest) set used image processing method, and ROI formed for each frame was received for continuous data to transmitter.
In this paper, three studies by using OCC-based V2I system were conducted. Firstly, the road signs were replaced by using Web server-client. The following scenario was used; The road sign was replaced by data packet which was received through the camera, the road sign information was requested from the web server. The server responded to road sign information. Finally, corresponding road sign information was displayed on the user''s GUI. For performance analysis, we measured the time taken from the image processing to displaying road signs on the GUI.
Secondly, in the case of the OCC used by the conventional image processing, the background of the visible light to be detected was important and if the sudden speed change occurred, the data reception rate was reduced due to the ambiguous region of ROI. This research explained how to improve the data reception rate by using the Deep Neural Network to detect ROI at every frame and to receive continuous data. For the analysis of performance, we compared the data reception rate for two conventional method and deep neural network method.
Finally, the technique of selecting three light sources to be used for localization in VLC-based V2I systems was proposed. In case of multiple LEDs in the real space the farthest point selection method, was employed to select the three most optimal LEDs for accurate positioning. After selecting the three optimal LEDs, the location was estimated through the collinality condition and the rotation conversion matrix. The simulation results showed that the proposed method was more accurate and faster to estimate the location them the method of the random selection of LEDs.
On the basis of experimental results, we believe that the intelligent transportation system can help to realize the autonomous driving system with high.