| Title |
Method of Detecting a Parking Violation Using Deep Running Tracking |
| Authors |
류지형(Jee-Hyung Rheu) ; 최두현(Doo-Hyun Choi) ; 김영모(Young-Mo Kim) |
| DOI |
https://doi.org/10.5573/ieie.2019.56.9.67 |
| Keywords |
Vehicle Tracking, Parking Violation, CNTK, Faster R-CNN, Backward-Sequence Video |
| Abstract |
This paper introduces an effective method for determining whether a driver has a parking violation. The method using the techniques based on a deep learning and LPR(License Plate Recognition) can obtain a plate number while tracking vehicles across ROI and obtain a reverse video including evidence images. A common way to crack down on the parking violations is to identify the car number from multiple images taken at the time of the parking violation. This method, which uses tracking and backward-sequence video, is more effective than the conventional method because it can recognize the license plate regardless of the license plate obscured by the object. In addition, license plate recognition rate is high because it obtains shot image at the best time to read the license plate. CNTK and Faster R-CNN were used in order to track vehicle and determine the parking violation. The experiment was conducted on 521 vehicles that passed through the road(Region Of Interest), of which the 30 vehicles were in violation of the parking law. By using the proposed method, 29 vehicles of them were judged to be parking violations. The detection rate of illegal vehicles was 96.7%. |