CONF
Yao_ECCV-VS_2008/IDIAP
Fast human detection from videos using covariance features
Yao, Jian
Odobez, Jean-Marc
EXTERNAL
https://publications.idiap.ch/attachments/papers/2008/Yao_ECCV-VS_2008.pdf
PUBLIC
https://publications.idiap.ch/index.php/publications/showcite/Yao_Idiap-RR-68-2007
Related documents
European Conference on Computer Vision, workshop on Visual
Surveillance (ECCV-VS)
Marseille
2008
October 2008
In this paper, we present a fast method to detect humans from videos captured in surveillance
applications.
It is based on a cascade of LogitBoost classifiers relying on features mapped from the Riemanian
manifold of region covariance matrices computed from input image features.
The method was extended in several ways. First, as the mapping process is slow for high
dimensional feature space, we propose to select weak classifiers based on subsets of
the complete image feature space.
In addition, we propose to combine these sub-matrix covariance features with the means of the image
features computed within the same subwindow, which are readily available from the covariance
extraction process.
Finally, in the context of video acquired with stationary cameras, we propose to fuse image
features from the spatial and temporal domains in order to jointly learn the correlation between
appearance and foreground information based on background subtraction.
Our method evaluated on a large set of videos coming from several databases (CAVIAR, PETS, ...,',','),
and can process from 5 to 20 frames/sec (for a 384x288 video) while achieving similar or better
performance than existing methods.
REPORT
Yao_Idiap-RR-68-2007/IDIAP
Fast Human Detection from Videos Using Covariance Features
Yao, Jian
Odobez, Jean-Marc
EXTERNAL
https://publications.idiap.ch/attachments/reports/2007/Yao_Idiap-RR-68-2007.pdf
PUBLIC
Idiap-RR-68-2007
2007
Idiap
December 2007
In this paper, we present a fast method to detect humans from videos captured in surveillance applications. It is based on a cascade of LogitBoost classifiers relying on features mapped from the Riemanian manifold of region covariance matrices computed from input image features. The method was extended in several ways. First, as the mapping process is slow for high dimensional input image feature space, we propose to select weak classifiers based on subsets of the complete image feature space, corresponding to sub-matrices of the full covariance matrix. In addition, we propose to combine these sub-matrix covariance features with the means of the image features computed within the same subwindow, which are readily available from the fast covariance extraction process based on integral images. Finally, in the context of video acquired with stationary cameras, we propose to fuse image features from the spatial and temporal domains in order to take advantage of both appearance and foreground information based on background subtraction to detect humans. We evaluated our method on a large dataset of videos coming from several databases (CAVIAR, PETS, ...). The results show that our approach can process from 5 to 20 frames/second (for a 384x288 video) while achieving similar performance than existing methods.