Master of Engineering Science
Electrical and Computer Engineering
Dr. Jagath Samarabandu
Safety has been a very crucial aspect in our lives and much attention has been paid to this issue since we need to remain safe everywhere.
In 2010, about 270,000 pedestrians were killed on the roads globally, which shows the importance of investigation of different approaches to reduce traffic fatalities.
One way to decrease the number of car accidents with pedestrians is to equip vehicles with cameras that detect and track pedestrians in the road. Many applications have been presented to improve the performance of pedestrian tracking.
However, it has remained a very challenging topic over the past few decades.
In this thesis, an automatic method is proposed for multiple pedestrian tracking.
State-of-the-art detection and tracking algorithms have been used in this study, followed by a novel post stage processing to increase the accuracy.
Proposed automatic tracking system was compared with a state-of-the-art tracking algorithm which shows comparable accuracy when used with the original incomplete ground truth data. It is estimated to offer better accuracy with a more accurate ground truth data.
The proposed algorithm offers potential improvements in both true positive ratio as well as false negative ratio when compared with the existing algorithm.
Our method is applicable in both outdoor applications such as tracking pedestrians that are walking in the street as well as indoor applications such as tracking people inside a building.
Ramin, Marjan, "Improvements to Tracking Pedestrians in Video Streams Using a Pre-trained Convolutional Neural Network" (2016). Electronic Thesis and Dissertation Repository. 3886.