Visual Tracking Using Deep Motion Features
Generic visual tracking is a challenging computer vision problem, where the position of a specified target is estimated through a sequence of frames. The only given information is the initial location of the target. Therefore, the tracker has to adapt and learn any kind of object, which it describes...
Main Author: | |
---|---|
Format: | Others |
Language: | English |
Published: |
Linköpings universitet, Datorseende
2016
|
Subjects: | |
Online Access: | http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-134342 |
id |
ndltd-UPSALLA1-oai-DiVA.org-liu-134342 |
---|---|
record_format |
oai_dc |
spelling |
ndltd-UPSALLA1-oai-DiVA.org-liu-1343422017-02-07T05:17:31ZVisual Tracking Using Deep Motion FeaturesengVisuell följning med hjälp av djup inlärning och optiskt flödeGladh, SusannaLinköpings universitet, Datorseende2016Visual trackingtrackingoptical flowdeep featuresDCFcorrelation filtersSRDCFcomputer visionGeneric visual tracking is a challenging computer vision problem, where the position of a specified target is estimated through a sequence of frames. The only given information is the initial location of the target. Therefore, the tracker has to adapt and learn any kind of object, which it describes through visual features used to differentiate target from background. Standard appearance features only capture momentary visual information. This master’s thesis investigates the use of deep features extracted through optical flow images processed in a deep convolutional network. The optical flow is calculated using two consecutive images, and thereby captures the dynamic nature of the scene. Results show that this information is complementary to the standard appearance features, and improves performance of the tracker. Deep features are typically very high dimensional. Employing dimensionality reduction can increase both the efficiency and performance of the tracker. As a second aim in this thesis, PCA and PLS were evaluated and compared. The evaluations show that the two methods are almost equal in performance, with PLS actually receiving slightly better score than the popular PCA. The final proposed tracker was evaluated on three challenging datasets, and was shown to outperform other state-of-the-art trackers. Student thesisinfo:eu-repo/semantics/bachelorThesistexthttp://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-134342application/pdfinfo:eu-repo/semantics/openAccess |
collection |
NDLTD |
language |
English |
format |
Others
|
sources |
NDLTD |
topic |
Visual tracking tracking optical flow deep features DCF correlation filters SRDCF computer vision |
spellingShingle |
Visual tracking tracking optical flow deep features DCF correlation filters SRDCF computer vision Gladh, Susanna Visual Tracking Using Deep Motion Features |
description |
Generic visual tracking is a challenging computer vision problem, where the position of a specified target is estimated through a sequence of frames. The only given information is the initial location of the target. Therefore, the tracker has to adapt and learn any kind of object, which it describes through visual features used to differentiate target from background. Standard appearance features only capture momentary visual information. This master’s thesis investigates the use of deep features extracted through optical flow images processed in a deep convolutional network. The optical flow is calculated using two consecutive images, and thereby captures the dynamic nature of the scene. Results show that this information is complementary to the standard appearance features, and improves performance of the tracker. Deep features are typically very high dimensional. Employing dimensionality reduction can increase both the efficiency and performance of the tracker. As a second aim in this thesis, PCA and PLS were evaluated and compared. The evaluations show that the two methods are almost equal in performance, with PLS actually receiving slightly better score than the popular PCA. The final proposed tracker was evaluated on three challenging datasets, and was shown to outperform other state-of-the-art trackers. |
author |
Gladh, Susanna |
author_facet |
Gladh, Susanna |
author_sort |
Gladh, Susanna |
title |
Visual Tracking Using Deep Motion Features |
title_short |
Visual Tracking Using Deep Motion Features |
title_full |
Visual Tracking Using Deep Motion Features |
title_fullStr |
Visual Tracking Using Deep Motion Features |
title_full_unstemmed |
Visual Tracking Using Deep Motion Features |
title_sort |
visual tracking using deep motion features |
publisher |
Linköpings universitet, Datorseende |
publishDate |
2016 |
url |
http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-134342 |
work_keys_str_mv |
AT gladhsusanna visualtrackingusingdeepmotionfeatures AT gladhsusanna visuellfoljningmedhjalpavdjupinlarningochoptisktflode |
_version_ |
1718412903561822208 |