Tracking Using Online Feature Selection and a Local Generative Model

Woodley, Thomas; Stenger, Björn; Cipolla, Roberto

doi:10.5244/c.21.86

Cited by 24 publications

(15 citation statements)

References 20 publications

Supporting

Mentioning

Contrasting

Unclassified

Order By: Relevance

“…Another extreme is Co-learning as in [19], where an online semi-supervised learning framework treated tracking as a foreground/background classification problem. Additionally, combinations of generative and discriminative models were used [20]. A boosted particle filter [4] and "detect to connect" [7,8] are of potential to deal with visual object tracking, because they first detect the objects of interest and then construct trajectories by analysis of motion continuity and appearance similarity.…”

Section: Related Workmentioning

confidence: 99%

Online Boosting-Based Object Tracking

Binh

2014

Proceedings of the 12th International Conference on Advances in Mobile Computing and Multimedia

View full text Add to dashboard Cite

Object tracking has been studied for years. The tasks of tracking (retrieving the object to be tracked) is still a challenging problem. In this paper, we present a novel and robust framework for automatic visual object tracking from image sequences or video. The main contribution is a new on-line boosting framework for efficient object tracking from image sequences. Boosting with interactive on-line training allows the object tracker to be trained and improve efficiently. The supervised classifiers are carefull trained on-line in order to increase adaptivity while limiting accumulation of error, i.e. drifting. a framework to solve the drifting and losing of tracking object during tracking. The main idea is to incorporate decision of given by a prior learned strong detector and an on-line adaptation tracking mechanism. The strong detector allows handling complex objects in even complicated environment. The cooperating framework exploits the power of a strong detector and a robust tracker allows to deal with drifting and losing effectively. In the experiments, we demontrate real-time tracking on several challenging sequences, including multi-object tracking of other objects. We outperform other on-line tracking methods especially in case of occlusions and presence of similar objects.

show abstract

Section: Related Workmentioning

confidence: 99%

Online Boosting-Based Object Tracking

Binh

2014

Proceedings of the 12th International Conference on Advances in Mobile Computing and Multimedia

View full text Add to dashboard Cite

show abstract

“…Therefore, Grabner [15] introduced on-line boosting to update feature weights to attain a compatibility of adaptation and stability of tracking classifiers. Woodley [16] employed discriminative feature selection using a local generative model to cope with appearance change while maintaining the proximity to a static appearance model.…”

Section: Introductionmentioning

confidence: 99%

Visual Tracking Algorithm Using Pixel-Pair Feature

Nishida

Kurita

Ogiuchi

et al. 2010

2010 20th International Conference on Pattern Recognition

View full text Add to dashboard Cite

A novel visual tracking algorithm is proposed in this paper. The algorithm uses pixel-pair features to discriminate between an image patch with an object in the correct position and image patches with an object in an incorrect position. The pixel-pair feature is considered to be robust for the illumination change, and also is robust for partial occlusion when appropriate features are selected in every video frame. The tracking precision for a deforming object (skier) is examined and also the occlusion detection method is described.

show abstract

“…Online adaptation of the measurement model is an ongoing area of study, e.g. [11,35], the main issue being the trade-off between adaptability versus drift. A time-weighted appearance model can be obtained through weighted online learning [36] that avoids drift and emphasizes most recent observations.…”

Section: Related Workmentioning

confidence: 99%

Motion Models that Only Work Sometimes

Cifuentes¹,

Sturzel²,

Jurie³

et al. 2012

Procedings of the British Machine Vision Conference 2012

View full text Add to dashboard Cite

It is too often that tracking algorithms lose track of interest points in image sequences. This persistent problem is difficult because the pixels around an interest point change in appearance or move in unpredictable ways. In this paper we explore how classifying videos into categories of camera motion improves the tracking of interest points, by selecting the right specialist motion model for each video. As a proof of concept, we enumerate a small set of simple categories of camera motion and implement their corresponding specialized motion models. We evaluate the strategy of predicting the most appropriate motion model for each test sequence. Within the framework of a standard Bayesian tracking formulation, we compare this strategy to two standard motion models. Our tests on challenging real-world sequences show a significant improvement in tracking robustness, achieved with different kinds of supervision at training time.frame 0 frame 10 frame 20 Figure 1: Example sequence with overlaid box showing the output of our specialized "forward" motion model, where the velocity and scale of objects approaching the camera tend to increase. Neither Brownian nor constant-velocity motion models are as successful at tracking interest points here.

show abstract

Tracking Using Online Feature Selection and a Local Generative Model

Cited by 24 publications

References 20 publications

Online Boosting-Based Object Tracking

Online Boosting-Based Object Tracking

Visual Tracking Algorithm Using Pixel-Pair Feature

Motion Models that Only Work Sometimes

Contact Info

Product

Resources

About