Reprinted from: http://blog.sciencenet.cn/blog-722391-571072.html
I have done some work on Background Modeling and moving object detection before. I plan to make a summary. Let's start with this cvpr2011 evaluation article. Evaluation
Of Background Subtraction Techniques for video surveillance (PDF)
Sebastian brutzer,
Benjamin hoeferlin (University of Stuttgart), Gunther heidemann (University of Stuttgart) project home page: http://www.vis.uni-stuttgart.de/index.php? Id = SABS can download the latest database and some evaluation code on this webpage (note that the evaluation code is not the code of the Background Modeling Method ). This article compares some background modeling methods in recent years. Comparison methods include:
I feel that the reason why this article can be published at a high-level meeting such as cvpr is as follows: 1. the author published a public data, and the database is a synthesis, so it is more convenient to quantify and evaluate other methods; 2. on the surface, the workload is huge. Although the author compared the 9 methods, almost all these methods have source code (integrated in opencv) on the Internet ), some other executable programs have been made public. The workload of the author is not great. We can see that the features column is basically characterized by color, while the author ignores the Background Modeling of texture features. Texture features: the University of ORU, who created the Background Modeling with texture, published the article on PAMI in. The author cannot help but understand it, I 'd like to ask him to compare the nine methods that are published on PAMI. Another 2010
Cvpr has an article on Stan Li, which is also made of textures. Both articles work well. I do not know why the author has not compared ...... we will introduce the two classic methods later. 3. the analysis is good. It seems that all cvpr articles have good analysis.
The author analyzed the following difficulties in Background Modeling:
- Gradual
Illumination changes: it is desirable that background model adapts to gradual changes of the appearance of the environment. For example in outdoor settings, the light intensity typically varies during day.
- Sudden illumination changes: sudden once-off changes are not covered by the background model. They occur for example with sudden switch of light, strongly
Affect the appearance of background, and cause False Positive detections.
- Dynamic Background: some parts of the scenery may contain movement, but shocould be regarded as background, according to their relevance. Such movement can
Be periodical or irregular (e.g., traffic lights, waving trees ).
- Camouflage: intentionally or not, some objects may poorly differ from the appearance of background, making correct classification difficult. This is especially
Important in surveillance applications.
- Shadows: shadows cast by foreground objects often complicate further processing steps subsequent to background subtraction. Overlapping shadows of foreground
Regions for example hinder their separation and classification. Hence, it is preferable to ignore these irrelevant regions.
- Bootstrapping: If initialization data which is free from foreground objects is not available, the background model has to be initialized using a bootstrapping
Strategy.
- Video noise: video signal is generally superimposed by noise. Background Subtraction approaches for video surveillance have to conflict with such degraded Signals
Affected by different types of noise, such as sensor noise or compression artifacts.
Evaluation Result:
It is worth noting that the barnich method has good speed and performance. His article contains pseudocode. The author's home page provides executable programs and can be integrated into his own programs.
