Road Sign Board Direction and Location Extraction and Recognition for Autonomous Vehicle.

Abstract

The problem of direction and location identification is very important in technologies used for Autonomous vehicles. while the navigation systems are that they cannot cover all areas due to a lack of signals or changes made on routes due to maintenance or upgrades. This research will focus on recognizing the sign and extracting address location names and directions from road signs. Moreover, it will help better identify road exits and lane directions for better route planning. In this paper we use YOLOv5 to identify the road board sign location and direction. Then extract the direction of each address location that are included in the road board sign and inform the car about the direction because autonomous car has no any driver so the car must decide by itself witch direction to choose to get the goal address location. This system can be used to continuously cheek the frames of the video that is taken by the car’s camera for road sign boards and analyses the image to find the direction of each location that are explained inside road sign board on the road. The proposed system consists of a camera mounted on top of the front mirror of the vehicle, and also a computer to run the recorded video on the system. In experiments, yolov5 framework achieves the best performance of 98.76% mean average precision (mAP) at Intersection over Union (IoU) threshold of 0.5, evaluated on our new developed dataset. And 91.31% on different IoU thresholds, ranging from 0.5 to 0.95.

Keywords

• Road Sign Board
• Autonomous vehicle
• YOLOv5