A Safety Vehicle Detection Mechanism Based on YOLOv5
Yeting Huang, Hancui Zhang
Abstract
Vehicle affect the safety of people's daily life, and it is difficult to detect and discriminate the surrounding conditions of the vehicle because of a variety of object types and different sizes. The paper aims to establish a digital assistant driving detection system, enhance the rapid detection of this multi-scale objects, and provide effective decision-making responses to ease the traffic accidents. The paper proposes a safety pre-warning detection method based on YOLOv5. Two mainly detection models YOLOv3 and YOLOv5 are taken to estimate the accuracy and efficiency of the detection mechanism. Experimental results show that YOLOv5 can obtain a higher average detection speed and be good for decision-making. Furthermore, in order to test the real driving data, a software system has been developed, and a real scenario application environment and some test experiments are conducted to prove the efficiency of the proposed mechanism.