Application of Vehicle Recognition Model Based on Yolo Neural Network on Embedded Devices

Abstract

With the development of the times, vehicles have become an indispensable means of transportation in people’s daily life. A large number of vehicles require scientific and effective detection and management. Existing vehicle recognition technologies are only used for computer training and operation, and cannot achieve good results in embedded devices. When the photographed vehicle angle is not good or the photographed license plate information is not detailed, the recognition effect is average. Compared with the workstation and other equipment, the vehicle recognition speed in the embedded device is slow. This paper proposes a vehicle detection and license plate character recognition model for an embedded platform and applies it to a real environment. yolo (you only watch it once) accomplishes vehicle detection by training a model on a workstation and deploying the trained network model to an embedded device. After training the vehicle detection model with a large dataset provided by the company, the yolo model has been able to detect vehicles very accurately. On this basis, the detected vehicle area is subjected to license plate detection and then the ocr character recognition is used to realize the license plate recognition. The model has been deployed to the embedded platform and has the characteristics of good timeliness and high accuracy. It can be applied to scenarios such as parking lots or high-speed patrol gates.