comparison of matrix completion algorithms for background initialization in videos
TRANSCRIPT
Comparison of Matrix Completion Algorithms for Background Initialization in Videos
Andrews Sobral, Thierry Bouwmans and El-hadi ZahZahPh.D. Student, Computer Vision
Lab. L3I/MIA – University of La Rochelle, France
Summary
▪ Context
▪ Understanding an Intelligent Video Surveillance Framework
▪ Introduction to Background Subtraction
▪ Background Model Initialization Problem
▪ Matrix Completion
▪ Proposed Approach
▪ Experimental results
▪ Conclusions
!
Video Content Analysis(VCA) or Video Analytics
Behavior Analysis
Image acquisition and preprocessing
ObjectDetection
ObjectTracking
event location
Intrusion detection Collision
prevention
Target detection and tracking
Anomaly detection
Target behavior analysis
Traffic data collection and
analysis
activity report
Understanding an Intelligent Video Surveillance Framework
supervisor
Behind the Scenes of an Intelligent Video Surveillance Framework
!
Example of automatic incident detection
our focus
Introduction to Background Subtraction
Initialize Background Model
frame modelForegroundDetection
Background Model Maintenance
our focus
Background subtraction methods
Traditional methods:• Basic methods, mean and variance over time• Fuzzy based methods• Statistical methods • Non-parametric methods• Neural and neuro-fuzzy methods
Matrix and Tensor Factorization methods:• Eigenspace-based methods (PCA / SVD)• RPCA, LRR, NMF, MC, ST, etc.• Tensor Decomposition, NTF, etc.
BGSLibrary (C++)https://github.com/andrewssobral/bgslibrary
A large number of algorithms have been proposed for background subtraction over the last few years:
LRSLibrary (MatLab)https://github.com/andrewssobral/lrslibrary
our focus
Andrews Sobral and Antoine Vacavant. A comprehensive review of background subtraction algorithms evaluated with synthetic and real videos. Computer Vision and Image Understanding (CVIU), 2014. http://dx.doi.org/10.1016/j.cviu.2013.12.005
Introduction to Matrix Completion (MC)
▪ MC can be formulated as the problem of o recover a low rank matrix (L) from the partial observations of its entries:
L
Underlying low-rank matrix
A
Matrix of partial observations
http://perception.csl.illinois.edu/matrix-rank/home.html
Motion Detection and Frame Selection
92 relevant frames are selected from a total of 296 frames (68,92% of reduction).
Matrix Completion Process
Illustration of the matrix completion process. From the left to the right: a) the selected frames in vectorized form (our observation matrix), b) the moving regions are represented by non-observed entries (black pixels), c) the moving regions filled with zeros (modified version of the observation matrix), and d) the recovered matrix after the matrix completion process.
Experimental results
Scene Background Initialization (SBI) data set
SBI data set provide 7 image sequences and corresponding ground truth backgrounds. Matlab scripts are provided for evaluating results in terms of eight metrics that include those used in the literature for background estimation.
Lucia Maddalena and Alfredo Petrosino, Towards Benchmarking Scene Background Initialization, arXiv:1506.04051http://sbmi2015.na.icar.cnr.it/
Matrix Completion Algorithms
The algorithms were implemented in MATLAB (R2014a) running on a laptop computer with Windows 7 Professional 64 bits, 2.7 GHz Core i7-3740QM
processor and 32Gb of RAM.
Conclusions
▪ The key idea is to eliminate the redundant frames, and consider its moving regions as non-observed values. This approach results in a matrix completion problem, and the background model can be recovered even with the presence of missing entries.
▪ The experimental results on the SBI data set shows the comparative evaluation of these recent methods, and highlights the good performance of LRGeomCG method over its direct competitors.
▪ Future research may concern to evaluate incremental and real-time approaches of matrix completion algorithm applied in streaming videos.
▪ MATLAB codes available: https://sites.google.com/site/mc4bmi/