Loading...

Tracking Based on Trajectory Information

Taheri Hanjani, Mohammad Javad | 2022

52 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 56051 (05)
  4. University: Sharif University of Technology
  5. Department: Electrical Engineering
  6. Advisor(s): Behroozi, Hamid; Mohammadzadeh, Hoda
  7. Abstract:
  8. Object tracking is one of the first, most basic and among the topics of interest in the field of computer vision. Nowadays, with the availability of high-quality and inexpensive video cameras and the expansion of neural networks, there has been a great interest in automatic video analysis using object tracking algorithms. However, many of the existing object tracking algorithms do frame-by-frame tracking using videos with high frame rates, which is not suitable for all locations that use surveillance cameras, because due to existing hardware limitations, the recorded videos are either kept for a limited period of time or are forcibly stored with low frame rates, which leads to the loss of a lot of information and so-called sparseness. In this research, a method for tracking, using past achievements in the field of tracking multiple targets and deep learning, based on suitable videos with the assumption of the problem, has been presented, which has been tried to overcome the problems in these types of videos to an acceptable extent. This method is based on the previous movements of people in the frames of these videos. In fact, first, by using many past researches in the field of tracking based on deep networks, these traveled paths were extracted and then, if pre-processing is needed, a simple processing was also done on these paths. Then, the approximate paths of the tracks in these scenes are determined and based on the distance of the targets to these paths, an adaptive threshold is considered that has the ability to preserve the identities. At the end, the tracked targets are compared with the real routes available and the capabilities of this tracker are shown. As an example, with the help of the tracker provided in this research, in one of the videos available at the university level, we have been able to achieve MOTA score equal to 0.53 and IDsw equal to 103 in the presence of 81 different identities without using external features
  9. Keywords:
  10. Tracking ; Multiobjective Tracking ; Object Tracking ; Motion Capture ; Path Information ; Computer Vision

 Digital Object List

 Bookmark

No TOC