발간년도 : [2021]
| 논문정보 |
|
| 논문명(한글) |
[Vol.16, No.1] A Machine Learning System for Crosswalk Detection |
|
| 논문투고자 |
Ju-Seok Park, Gyesik Lee |
|
| 논문내용 |
There are some places where it is difficult for the blind to judge because of different types of braille blocks, mainly caused by broken braille blocks or the reinstallation of braille blocks due to rapid road revisions. Therefore braille blocks installed on crosswalks often cause accidents for the blind. To prevent this, the crosswalk has a walking signal voice guidance system and auxiliary devices. There have been various attempts to install such systems to prevent accidents. But there are many uninstalled places, or installed devices often don’t work properly, making it difficult to use them in practice. We would like to present a crosswalk detection algorithm using Deep Learning. If the algorithm is mounted on a small device, it can be attached to the user's body to prevent unexpected situations, unlike systems that need to be installed on all crosswalks. It is economical and can effectively prevent accidents by alerting the user. The crosswalk detection algorithm outputs sound when it detects a crosswalk, indicating that there is a crosswalk ahead of the user. To this end, the object detection algorithm YOLOv3 and the newral network framework were used. In this project, the model we trained learned from images of crosswalks taken directly and showed an average recognition rate of 92% or higher. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
