Download the full-sized PDF of Computing Velocity of Multiple Objects in Sequences of Images With an Application In Water-Based Bitumen Extraction ProcessDownload the full-sized PDF



Permanent link (DOI):


Export to: EndNote  |  Zotero  |  Mendeley


This file is in the following communities:

Graduate Studies and Research, Faculty of


This file is in the following collections:

Theses and Dissertations

Computing Velocity of Multiple Objects in Sequences of Images With an Application In Water-Based Bitumen Extraction Process Open Access


Other title
Tracking Multiple Objects
Detecting Multiple Objects
Recursive Iterative Thresholding Framework
Using Tracklets in Flow Network
Finding K-Shortest Path in Flow Network
Type of item
Degree grantor
University of Alberta
Author or creator
Shooshtari, Mahdi
Supervisor and department
Ray, Nilanjan (Computing Science)
Zhang, Hong (Computing Science)
Examining committee member and department
Boulanger, Pierre (Computing Science)
Department of Computing Science

Date accepted
Graduation date
2017-06:Spring 2017
Master of Science
Degree level
Image-based analysis of bitumen extraction process can provide the oil companies with useful information that can be used to assess their performance in retrieving bitumen from the oil sands. In this analysis, several slurry images are taken during the extraction process, and then image processing techniques are used to extract the information and operating metrics from the images. Computing the velocity of floating objects, including bitumen droplets and sand particles, is one of the most important operating metrics. This operating metric gives us information about bitumen retrieving performance in addition to more understanding about the floating objects and their movement; yet up to our knowledge, there is not an evaluated automated method to detect and track the floating objects in slurry images successfully and measure their speed in the real-time. To address this problem, we have developed algorithms to (1) detect bitumen droplets and sand particles in each image; and (2) track the detected objects in the sequence of images. For the first step of this project, we evaluated several well-known global and local thresholding and segmentation algorithms and found the method with the best outcome for our images. We also applied tiling and downsampling methods to the images in order to improve the performance of evaluated thresholding and segmentation algorithms in the object detection and/or decreasing running time of the segmentation algorithms. The results indicated that tiling and downsampling decrease the performance in most of the cases. Tiling and downsampling of the images decrease the running time of the segmentation algorithms, but they still are not as fast as thresholding methods. Moreover, we developed a Recursive Iterative Thresholding framework that can be combined with any local or global thresholding algorithm to improve the performance of the original algorithm through detecting a larger number of small and bright objects. For the second step of the project (i.e. tracking the detected objects in a sequence of images), we implemented and tested two tracking methods: (1) a frame-by-frame tracking algorithm; and (2) a multi-frame tracking method. For the frame-by-frame tracking algorithm, objects of consecutive images were matched and connected to each other by using an assigning algorithm that determines the optimum assignment. In contrast, for the multi-frame tracking method, multiple frames were considered together, and a Flow Network was built by connecting the objects using the weighted edges. Then a greedy approach was used to detect the K-shortest path (KSP) in the Flow Network. Our experiments indicated that finding the KSP in a Flow Network has better performance compared to the frame-by-frame tracking, and it produces the results in a noticeably less running time. We used two State-of-the-Art cutting algorithms and developed two additional novel methods to cut the tracks into smaller tracklets for using in the Flow Network. Our experiments showed that using each tracklet as a node of the Flow Network instead of using each detected object as a node improved the performance of the algorithm.
This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for the purpose of private, scholarly or scientific research. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.
Citation for previous publication

File Details

Date Uploaded
Date Modified
Audit Status
Audits have not yet been run on this file.
File format: pdf (PDF/A)
Mime type: application/pdf
File size: 1463088
Last modified: 2017:06:13 12:30:28-06:00
Filename: Shooshtari_Mahdi_201701_MSc.pdf
Original checksum: f0c99f944135850126638368f1c344c6
Well formed: true
Valid: true
Page count: 84
Activity of users you follow
User Activity Date