Objectness-based smoothing stochastic sampling and coherence approximate nearest neighbor for visual tracking

Mbelwa, Jimmy T.; Zhao, Qingjie; Lu, Yao; Líu, Hao; Wang, Fasheng; Mbise, Mercy

doi:10.1007/s00371-018-1470-5

Cited by 6 publications

(1 citation statement)

References 41 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Zhou et al [31] utilized nearest neighbor field estimation to compute the importance proposal probabilities, which guide the Markov chain search towards promising regions. Mbelwa et al [32] integrated prior knowledge and objectness proposal into the smoothing stochastic approximate Monte Carlo to predict abrupt motion. In this work, we utilize the DOS distribution to propose candidate regions that may contain targets, and introduce MSML to further improve the reliability of region proposal strategy.…”

Section: A Stochastic Sampling Based Tracking Methodsmentioning

confidence: 99%

Uncertain Motion Tracking Combined Markov Chain Monte Carlo and Correlation Filters

et al. 2019

View full text Add to dashboard Cite

To address the uncertain motion tracking problem, a tracking method based on the Markov Chain Monte Carlo and correlation filters is proposed. Firstly, multi-scope marginal likelihood (MSML) strategy is introduced to Wang-Landau Monte Carlo (WLMC) tracking method for increasing the acceptance ratio of samples in the promising regions and obtaining a more reliable distribution of density-of-states (DOS). Secondly, in order to raise the efficiency of the tracker, DOS is used to mark the region of interest. Then correlation filters are used to simplify the iterative optimizing operation of the subregions, and eventually target positioning is achieved by maximum response in the promising regions. Finally, a unified tracking framework is designed to enable correlation filters and WLMC with MSML strategy to exploit and complement each other to cope with uncertain motion tracking. Extensive experimental results on uncertain Motion sequences and benchmark datasets demonstrate that the proposed method performs favorably against the state-of-the-art methods. INDEX TERMS Markov chain Monte Carlo, correlation filters, uncertain motion, visual tracking. I. INTRODUCTION Visual tracking is an active research topic in computer vision community that finds numerous applications such as intelligent surveillance, medical research, motion analysis, and autonomous driving [1]-[3]. In the past decades, dramatic progress has been achieved [4], but there are still some very challenging and unresolved issues, including illumination changes, fast motions, pose variations, partial occlusions and background clutters and so on. In order to obtain better tracking results, many trackers based on deep learning or correlation filtering have been proposed in recent years. Examples include target-aware deep tracking [5], multi-task correlation particle filter [6], hierarchical convolutional features [7], adaptive hedging [8], spatial regularization [9], continuous convolutional operations [10], etc. Despite the great success of the above methods in visual tracking, most existing approaches are usually based on a smooth motion assumption. In real-word scenarios, it is The associate editor coordinating the review of this manuscript and approving it for publication was Shiping Wen .

show abstract

Section: A Stochastic Sampling Based Tracking Methodsmentioning

confidence: 99%