English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Tracking a target of interest in both sparse and crowded environments is achallenging problem, not yet successfully addressed in the literature. In thispaper, we propose a new long-term visual tracking algorithm, learningdiscriminative correlation filters and using an online classifier, to track atarget of interest in both sparse and crowded video sequences. First, we learna translation correlation filter using a multi-layer hybrid of convolutionalneural networks (CNN) and traditional hand-crafted features. We combineadvantages of both the lower convolutional layer which retains more spatialdetails for precise localization and the higher convolutional layer whichencodes semantic information for handling appearance variations, and thenintegrate these with histogram of oriented gradients (HOG) and color-namingtraditional features. Second, we include a re-detection module for overcomingtracking failures due to long-term occlusions by training an incremental(online) SVM on the most confident frames using hand-engineered features. Thisre-detection module is activated only when the correlation response of theobject is below some pre-defined threshold. This generates high score detectionproposals which are temporally filtered using a Gaussian mixture probabilityhypothesis density (GM-PHD) filter to find the detection proposal with themaximum weight as the target state estimate by removing the other detectionproposals as clutter. Finally, we learn a scale correlation filter forestimating the scale of a target by constructing a target pyramid around theestimated or re-detected position using the HOG features. We carry outextensive experiments on both sparse and dense data sets which show that ourmethod significantly outperforms state-of-the-art methods.

Long-term Correlation Tracking using Multi-layer Hybrid Features in Sparse and Dense Environments