Multi-object tracking in video using labeled random finite sets

Rathnayake, T 2018, Multi-object tracking in video using labeled random finite sets, Doctor of Philosophy (PhD), Engineering, RMIT University.


Document type: Thesis
Collection: Theses

Attached Files
Name Description MIMEType Size
Chapter_2_Tracking_Results_CAVIAR.zip Video 16 sec Click to show the corresponding preview/stream application/zip 20.83MB
Chapter_4_Tracking_Results_Sequence1.zip Video 11 sec Click to show the corresponding preview/stream application/zip 6.20MB
Chapter_4_Tracking_Results_Sequence2.zip Video 27 sec Click to show the corresponding preview/stream application/zip 15.48MB
Chapter_4_Tracking_Results_Sequence3.zip Video 15 sec Click to show the corresponding preview/stream application/zip 7.72MB
Chapter_5_Tracking_Results_ETH_Bahnhof.mp4 Video 33 sec Click to show the corresponding preview/stream video/x-m4v 38.93MB
Chapter_5_Tracking_Results_ETH_SunnyDay.mp4 Video 12 sec Click to show the corresponding preview/stream video/x-m4v 12.20MB
Chapter_5_Tracking_Results_PETS2009S2L1V1.mp4 Video 26 sec Click to show the corresponding preview/stream video/x-m4v 20.94MB
Chapter_5_Tracking_Results_TUD_Stadtmitte.mp4 Video 6 sec Click to show the corresponding preview/stream video/x-m4v 4.28MB
Chapter_6_Tracking_Results_S2L1.mp4 Video 26 sec Click to show the corresponding preview/stream video/mp4 4.90MB
Chapter_6_Tracking_Results_S2L1_3D.mp4 Video 26 sec Click to show the corresponding preview/stream video/mp4 4.94MB
Chapter_6_Tracking_Results_S2L2.mp4 Video 14 sec Click to show the corresponding preview/stream video/mp4 2.60MB
Chapter_6_Tracking_Results_S2L3.mp4 Video 7 sec Click to show the corresponding preview/stream video/mp4 1.72MB
Rathnayake.pdf Thesis application/pdf 36.60MB
Title Multi-object tracking in video using labeled random finite sets
Author(s) Rathnayake, T
Year 2018
Abstract The safety of industrial mobile platforms (such as fork lifts and boom lifts) is of major concern in the world today as industry embraces the concepts of Industry 4.0. The existing safety methods are predominantly based on Radio Frequency Identification (RFID) technology and therefore can only determine the distance at which a pedestrian who is wearing an RFID tag is standing. Other methods use expensive laser scanners to map the surrounding and warn the driver accordingly. The aim of this research project is to improve the safety of industrial mobile platforms, by detecting and tracking pedestrians in the path of the mobile platform, using readily available cheap camera modules.

In order to achieve this aim, this research focuses on multi-object tracking which is one of the most ubiquitously addressed problems in the field of \textit{Computer Vision}. Algorithms that can track targets under severe conditions, such as varying number of objects, occlusion, illumination changes and abrupt movements of the objects are investigated in this research project. Furthermore, a substantial focus is given to improving the accuracy and, performance and to handling misdetections and false alarms. In order to formulate these algorithms, the recently introduced concept of Random Finite Sets (RFS) is used as the underlying mathematical framework.

The algorithms formulated to meet the above criteria were tested on standard visual tracking datasets as well as on a dataset which was created by our research group, for performance and accuracy using standard performance and accuracy metrics that are widely used in the computer vision literature. These results were compared with numerous state-of-the-art methods and are shown to outperform or perform favourably in terms of the metrics mentioned above.
Degree Doctor of Philosophy (PhD)
Institution RMIT University
School, Department or Centre Engineering
Subjects Computer Vision
Signal Processing
Image Processing
Keyword(s) Multi-object tracking
Multi-target tracking
Visual tracking
Random finite sets
Bernoulli
Bayesian
Computer vision
Signal processing
Image processing
Information fusion
Industrial safety
Versions
Version Filter Type
Access Statistics: 24 Abstract Views, 102 File Downloads  -  Detailed Statistics
Created: Thu, 29 Nov 2018, 09:42:19 EST by Keely Chapman
© 2014 RMIT Research Repository • Powered by Fez SoftwareContact us