SteveScher's tracking [4 articles]

Abstract

DOTS (Dynamic Object Tracking System) is an indoor, real-time, multi-camera surveillance system, deployed in a real office setting. DOTS combines video analysis and user interface components to enable security personnel to effectively monitor views of interest and to perform tasks such as tracking a person. The video analysis component performs feature-level foreground segmentation with reliable results even under complex conditions. It incorporates an efficient greedy-search approach for tracking multiple people through occlusion and combines results from individual cameras into multi-camera trajectories. The user interface draws the users’ attention to important events ...

Note (first note only)

Example of a reasonably complete and complicated system of cameras for surveillance and monitoring. Not all that much technical detail or theoretical contribution, but a neat system and an example of demand for this.

Abstract

In this paper, we present a novel view to home video content analysis, which aims at tracking the capture intention of camcorder users. Based on the study of intention mechanism in psychology, a set of domain-specific capture intention concepts are defined. A comprehensive and extensible scheme consisting of video structuring, intention oriented feature analysis, as well as intention unit segmentation and classification is proposed to mine the users' capture intention. Experiments were carried on home video sequences of 90 ...

Abstract

Non-rigid object detection and articulated pose estimationare two related and challenging problems in computervision. Numerous models have been proposed over theyears and often address different special cases, such aspedestrian detection or upper body pose estimation in TVfootage. This paper shows that such specialization may notbe necessary, and proposes a generic approach based onthe pictorial structures framework. We show that the rightselection of components for both appearance and spatialmodeling is crucial for general applicability and overallperformance of the model. The appearance of ...

Abstract

In this paper, we address the problem of learning anadaptive appearance model for object tracking. In particular,a class of tracking techniques called “tracking by detection”have been shown to give promising results at realtimespeeds. These methods train a discriminative classifierin an online manner to separate the object from the background.This classifier bootstraps itself by using the currenttracker state to extract positive and negative examplesfrom the current frame. Slight inaccuracies in the trackercan therefore lead to incorrectly labeled training examples,which degrades the classifier ...

Note (first note only)

I like Multiple-Instance Learning, I should read the original MIL papers. MIL addresses the problem where at least one of the "instances" (e.g. photos, feature vectors) in a group ("bag") is known to be a true positive, but not which one. Seems like a good way to model uncertainty in some situations.

This paper departs from that basic notion a few times. It uses boosting, and trains each weak classifier under the assumption that all instances in the bag are positive. I