Hu, Xizhi
(School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou, China)
,
Wei, Zheng
(School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou, China)
,
Zhou, Wenchao
(School of Mechanical and Automotive Engineering, South China University of Technology, Guangzhou, China)
To achieve safe driving of vehicles, it is necessary to perceive information about the vehicle's surroundings, and computer vision is one of the key technologies to solve this problem. The YOLO series and SSD, RetinaNet algorithm are representative of one-stage target detection algorithms, which hav...
To achieve safe driving of vehicles, it is necessary to perceive information about the vehicle's surroundings, and computer vision is one of the key technologies to solve this problem. The YOLO series and SSD, RetinaNet algorithm are representative of one-stage target detection algorithms, which have high accuracy and high speed. YOLOv4 is the latest algorithm of YOLO series, which has improved the speed and accuracy of vehicle target detection than before, but there is still a distance from the real real-time in vehicle detection. This paper proposes an improved YOLOv4-based video stream vehicle target detection algorithm to solve the problem in the detection speed which is not fast enough. This paper first introduces the YOLOv4 algorithm theoretically, then proposes an algorithmic process to speed up the detection speed, and finally conducts practical road experiments. From the experimental results, the algorithm of this paper can improve the detection speed of the algorithm without losing accuracy, which can provide a basis for decision making for safe vehicle driving.
To achieve safe driving of vehicles, it is necessary to perceive information about the vehicle's surroundings, and computer vision is one of the key technologies to solve this problem. The YOLO series and SSD, RetinaNet algorithm are representative of one-stage target detection algorithms, which have high accuracy and high speed. YOLOv4 is the latest algorithm of YOLO series, which has improved the speed and accuracy of vehicle target detection than before, but there is still a distance from the real real-time in vehicle detection. This paper proposes an improved YOLOv4-based video stream vehicle target detection algorithm to solve the problem in the detection speed which is not fast enough. This paper first introduces the YOLOv4 algorithm theoretically, then proposes an algorithmic process to speed up the detection speed, and finally conducts practical road experiments. From the experimental results, the algorithm of this paper can improve the detection speed of the algorithm without losing accuracy, which can provide a basis for decision making for safe vehicle driving.
※ AI-Helper는 부적절한 답변을 할 수 있습니다.