Electric kickboard safety regulation violation checking system using image classifier
Abstract
Author
Yun
Date
2021-07-09 16:24
Views
166
| Type of Presentation | Oral ( 0 ) / Poster ( ) / Anything ( ) |
| Scope and Interests | deep learning , machine learning |
| Title of Paper | Electric kickboard safety regulation violation checking system using image classifier |
|
Corresponding Author(s) |
Name: Yoosoo Oh Affiliation: Dept. of ICT Convergence, Dept. of AI, Daegu University, Gyeongsan-si, Republic of Korea e-mail: yoosoo.oh@daegu.ac.kr Tel:+82 53-850-6654 |
| Author(s) name / Affiliation / e-mail |
1) SeongU Yun 2)Dept. of ICT Convergence, Dept. of AI, Daegu University, Gyeongsan-si, Republic of Korea 3)kkk4561@daegu.ac.kr |
| Abstract | As the safety accidents of electric scooters have been increasing, the electric kickboard-related legislation was revised in 2020. The revised road traffic law about kickboards is mandatory to wear helmets and bans the onboarding of any passenger when driving electric kickboards. In this paper, we propose a system for checking violations of safety regulations for electric kickboards using an image classifier. First, the proposed violation enforcement system is installed where shared kickboards are frequently parked and stopped. Also, when driving an electric kickboard, our system determines whether the driver has worn a helmet and whether the passenger is on board or not. Then, our system emits a warning sound when the law is violated. The proposed system consists of MCU, camera, and buzzer. The proposed system implements the OpenCV to receive real-time video from the camera connected to the MCU and then estimates the violation of the law using the YOLO algorithm. In this paper, we compare and analyze the fastest v5s and the most accurate v5x among the four types of YOLO v5s, v5m, v5l, and v5x, and we adapt the exact algorithm. Our system determines whether the recognized person wears a helmet with the object recognition after receiving and learning the helmet/head dataset. We then establish a bounding box for the total size of objects recognized as human beings and objects recognized as electric kickboards to determine whether a passenger is on board or not. If our system recognizes two or more persons within the bounding box area of the electric kickboard, it estimates a passenger with our violation policy. Also, the proposed system makes an alarm when an electric kickboard driver is not wearing a helmet, or more than one person is on board. Accordingly, the proposed system encourages that electric kickboard drivers drive in a safe state. Therefore, the proposed approach is expected to reduce injury in an accident by wearing a helmet and reducing the risk of an accident by riding a single person. |
| Keywords | Yolo, kickboard, detect , MCU |