Deep Learning based Pedestrian Detection and Tracking System using Unmanned Aerial Vehicle and Prediction Method
Jae Hee Lee1, Chang Jin Seo2
1Jae Hee Lee, Department of Information & Communication, Dong Seoul University, Sungnam, South Korea, East Asian.
2Chang Jin Seo, Department of Information Security Engineering, Sangm yung University, Cheonan, South Korea, East Asian.
Manuscript received on 10 June 2019 | Revised Manuscript received on 17 June 2019 | Manuscript Published on 22 June 2019 | PP: 794-799 | Volume-8 Issue-8S2 June 2019 | Retrieval Number: H11330688S219/19©BEIESP
Open Access | Editorial and Publishing Policies | Cite | Mendeley | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open-access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Pedestrian detection and tracking system have become an essential field in the object detection and target tracking research area. This study proposes for developing and implementing the fast pedestrian detection and tracking system using Deep learning (YOLOv3), UAV (Unmanned Aerial Vehicle) and prediction method that is the Kalman Filter. Methods/Statistical analysis: The performance of the object detection and tracking system is decided by the performance time and the accuracy of object detection and tracking algorithms. So we applied to the YOLOv3 which is the fast detection method recent at our proposed system and also proposed the Kalman Filter algorithm with a variable detection area as the pedestrian tracking system. Findings: In the experiments, the proposed method successfully detected and tracked pedestrians who move at 53 FPS maximum and 38-43 FPS on average each test videos. The proposed way with variable search ranges made much fewer errors than the traditional object detector with fixed search ranges. Improvements/Applications: This research result can be applied pedestrian moving monitoring system, ITS (intelligent transport system) and security surveillance system.
Keywords: Pedestrian Tracking, Deep Learning, Object Detection, YOLOv3, Kalman Filter.
Scope of the Article: Deep Learning