Background and foreground modeling using nonparametric kernel density estimation for visual surveillance

By Ahmed M. Elgammal, Ramani Duraiswami, David Harwood, Larry S. Davis. Proceedings of the IEEE, 2002

1,331 citations recorded by the source corpus.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

View Paper: https://www.semanticscholar.org/paper/7ab4d9e7f70528a1cf43ecc8e3fc048a1b2fe0be

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

In Computer Vision and Pattern Recognition

Citations in the corpus

Object tracking: A survey (2006, background): “The probability density estimates of the object appearance can either be parametric, such as Gaussian [Zhu and Yuille 1996] and a mixture of Gaussians [Paragios and Deriche 2002], or nonparametric, such as Parzen windows [Elgammal et al. 2002] and histograms [Comaniciu et al.”
A texture-based method for modeling the background and detecting moving objects (2006, methodology): “For our method, we chose the local binary pattern (LBP) texture operator [2], [3], which has recently shown excellent performance in many applications and has several properties that favor its usage in background modeling.”
ViBe: A Universal Background Subtraction Algorithm for Video Sequences (2011, positive): “One of the strengths of nonparametric kernel density estimation methods [53], [55]–[59] is their ability to circumvent a part of the delicate parameter estimation step due to the”
Modeling pixel process with scale invariant local patterns for background subtraction in complex scenes (2010, methodology): “As a result, traditional numerical value based methods, either parametric one like Gaussian mixture models [8], or nonparametric one like kernel density estimation (KDE) [ 1 ], can not be used directly for modeling local patterns into background.”
Image change detection algorithms: a systematic survey (2005, methodology): “[60] observed that the coordinates ( G R+G+B , B R+G+B , R+ G + B) are more effective than RGB coordinates for suppressing unwanted changes due to shadows of objects.”
Moving Object Detection in Spatial Domain using Background Removal Techniques - State-of-Art (2008, methodology): “[2] use s a measure of lightness at each pixel.”
Beyond 5G With UAVs: Foundations of a 3D Wireless Cellular Network (2018, methodology): “To remove the discontinuity of cubes in the space, we use Gaussian kernels [33].”
Spatiotemporal Saliency in Dynamic Scenes (2010, methodology): “1. static cameras [ 12 ], [21], [25]; 2. explicit [14], or approximate [23], compensation of camera motion; 3. foreground objects that move in a consistent direction (an assumption that we denote as temporal coherence) [6], [16], [28] or have faster variations in appearance than the background [24];”
Changedetection.net: A new change detection benchmark dataset (2012, methodology): “[8] and GMM by Stauffer and Grimson [24], as well as 5 improved versions of these methods: self-adapting GMM by KaewTraKulPong [12], improved GMM by Zivkovic and van der Heijden [28], block-based GMM by Dora (RECTGAUSS-Tex) et al.”
A Review of Human Activity Recognition Methods (2015, methodology): “Background subtraction (Elgammal et al., 2002) and human tracking (Liu et al., 2010) are usually used as a part of this process; (ii) human activity modeling (e.g., feature extraction; Laptev, 2005) is the step of extracting the necessary information that will help in the recognition step; and…”
Learning Unsupervised Video Object Segmentation Through Visual Attention (2019, background): “…(UVOS), i.e ., automatically segmenting primary object regions from the background in videos, has been a long standing research challenge in computer vision [29, 30, 12, 23], and has shown potential beneﬁts for numerous applications, e.g ., action recognition [62] and object tracking [50].”
Learning object motion patterns for anomaly detection and improved object detection (2008, background): “There are certain components in traditional background subtraction algorithms [16, 3 ] that could benefit from this scene knowledge.”
Video Segmentation by Non-Local Consensus voting (2014, background): “These were assumed to be either: (i) stationary backgrounds [5, 9, 12, 33] or backgrounds undergoing 2D parametric motion [16, 18, 30] (thus image registration can be used for background stabilization); or (ii) backgrounds undergoing 3D motions with 3D parallax [17, 34, 37].”