발간년도 : [2023]
| 논문정보 |
|
| 논문명(한글) |
[Vol.18, No.6] Deep Learning Based Traffic Vehicle Image Classification System |
|
| 논문투고자 |
Jinwoo Park, Sangmin Suh |
|
| 논문내용 |
When a vehicle passes under a bridge or tries to enter a parking lot, the height of the vehicleprevents it from entering and often results in damage to the vehicle. To prevent this, traffic signs areoften used to alert drivers by indicating the height of vehicles that can enter the area. However, driversmay not see the signs due to poor road maintenance and poor visibility. Therefore, a device thatautomatically generates information about these height restrictions is needed, which can be obtainedautomatically by automatically classifying vehicle information. In this paper, we aim to prevent accidentson bridges and parking lots by automatically identifying the height of vehicles by automaticallyclassifying vehicle information. We designed a deep learning model based on convolutional neuralnetwork (CNN) for automatic classification of vehicle models. We used TensorFlow as the designframework and created three classification classes: bus, passenger car, and motorcycle. The dataset usedwas a total of 1500 images. For performance evaluation, we first calculated the chaos matrix and usedit to calculate precision and recall, and finally calculated the performance metric, F1-score. F1-score isthe fairest performance metric and we obtained 0.964, 0.902, and 0.926 for bus, car, and motorcycle,respectively, with an average performance of 0.933. |
|
| 첨부논문 |
|
|
|
|
|
자료실 > 논문지 
