Study

Visible Light Communication

In this study, an ROI detection method is proposed based on CNN, which is robust to changes in the environment and can improve ROI detection performance.

YOLO V3 Neural Network [YOLO V3 Reference Paper] is utilized to ensure real-time and accuracy.

The image information obtained through the camera from the moving vehicle is compared with the existing image processing method by selecting the ROI based on the coordinates output through the neural network in which the traffic lights are learned.

In addition, considering the ROI information of the previous frame, a threshold setting technique is proposed that enables adaptive changes to the environmental changes around the vehicle.

Actual driving tests have confirmed that ROI detection and threshold setting techniques using deep neural network techniques proposed in this study provide an improvement in V2I communication reception over conventional techniques.