Abstract | ||
---|---|---|
Object tracking, especially human tracking is one of the challenging research problems in computer vision. Although the performance has gained some positive changes recently, there is still room for improvement. In this paper, we introduce an approach for human detection and tracking using Convolution Neural Network (CNN) and Hungarian Algorithm (HA). A CNN is used to localize multiple human beings from frame to frame in a video stream. This deep CNN is known as Faster R-CNN which achieved the state of the art performance in object detection problem. In the tracking process, we solve the data association problem in visual tracking using HA. A detected person will be assigned to a tracklet based on the data distribution in the video frame. The experimental results show that our system can deal with the videos captured from different scenarios in near real-time. |
Year | DOI | Venue |
---|---|---|
2019 | 10.1007/s11042-018-6141-z | Multimedia Tools and Applications |
Keywords | Field | DocType |
Drone images, Faster R-CNN, Human tracking, Hungarian algorithm | Hungarian algorithm,Computer vision,Object detection,Pattern recognition,Computer science,Convolutional neural network,Eye tracking,Data association,Video tracking,Drone,Artificial intelligence | Journal |
Volume | Issue | ISSN |
78.0 | 4 | 1573-7721 |
Citations | PageRank | References |
1 | 0.39 | 14 |
Authors | ||
6 |
Name | Order | Citations | PageRank |
---|---|---|---|
Hai Duong Nguyen | 1 | 5 | 2.82 |
In Seop Na | 2 | 42 | 13.83 |
Soo-Hyung Kim | 3 | 191 | 49.03 |
Guee Sang Lee | 4 | 8 | 2.24 |
Hyungjeong Yang | 5 | 455 | 47.05 |
Junho Choi | 6 | 366 | 60.87 |