Abstract:Tracking multiple people under occlusion and across cameras is a challenging question for discussion. Furthermore, the cameras in this study are used to extend the field of view, which are distinguished from the same field of view. Such correspondence between multiple cameras is a burgeoning research subject in the area of computer vision. This paper effectively solves the problems of tracking multiple people who pass from one camera to another and segmenting people under occlusion using probabilistic models. The probabilistic models are composed of blob model, motion model and color model, which make the most of the space, motion and color information. First, we present a color model that uses maximum likelihood estimation based on non-parametric kernel density estimation. Second, we introduce a blob model based on mean shift, which segments the body into many regions according to the color of each person in order to spatially localize the color features corresponding to the way people are dressed. Clothes can be any mixture of colors. Third, we bring forward a motion model based on statistical probability which indicates the movement position of the same person between two successive frames in a single camera. Finally, we effectively unify the three models into a general probabilistic model and attain a maximization likelihood probability image, which is used to segment the foreground region under occlusion and to match people across multiple cameras.

On Exploiting Occlusions in Multiple-View Geometry

Tracking Multiple People under Occlusion and Across Cameras Using Probabilistic Models

Multi-View Geometry Compression

TC-SfM: Robust Track-Community-Based Structure-from-Motion

Two-View Multibody Structure from Motion

Occlusion-aware Visual Tracker using Spatial Structural Information and Dominant Features

Finding Multiple Object Instances with Occlusion

Occlusion Boundary: A Formal Definition & Its Detection Via Deep Exploration of Context

A new geometric approach for three view line reconstruction and motion estimation in Manhattan Scenes

An Anti-Occlusion Vision-Based Method for Structural Motion Estimation

Multiple Human Association and Tracking from Egocentric and Complementary Top Views

OSMO: Online Specific Models for Occlusion in Multiple Object Tracking under Surveillance Scene.

A Graph-Matching Approach for Cross-view Registration of Over-view 2 and Street-view based Point Clouds

Imagining the unseen

Time-based Mapping of Space Using Visual Motion Invariants

WALT3D: Generating Realistic Training Data from Time-Lapse Imagery for Reconstructing Dynamic Objects under Occlusion

Self-supervised Learning of Monocular 3D Geometry Understanding with Two- and Three-View Geometric Constraints

Trifocal Relative Pose from Lines at Points and its Efficient Solution

Occ$^2$Net: Robust Image Matching Based on 3D Occupancy Estimation for Occluded Regions

Generalized Rank Conditions In Multiple View Geometry With Applications To Dynamical Scenes

Imagining the Unseen: Stability-Based Cuboid Arrangements for Scene Understanding.