Vehicle Taillamp Intention Recognition for Intelligent and Connected Vehicles Based on YOLOv4
Bingming Tong, Luyao Du, Wei Chen, Hongjiang Zheng · 2021 4th International Conference on Advanced Electronic Materials, Computers and Software Engineering (AEMCSE) · 2021
In order to achieve fast and accurate vehicle taillamp intention recognition for intelligent and connected vehicles, a taillamp intention recognition method based on the YOLOv4 algorithm was proposed. Firstly, the images of the vehicles in different lamping conditions during driving were collected, and the corresponding dataset was established for training and testing the model. Then the input parameter was modified, and the cosine annealing learning rate algorithm was applied to the YOLOv4 framework. Experimental result showed that this method had an accuracy of 90.27% for vehicle taillamp intention recognition and a detection rate of 20fps. This method can realize the intention recognition of vehicle taillamps, and provide a basis for driving decision-making of intelligent and connected vehicles.