Moving target detection algorithm based on local two-value similarity mode (LBSP)
[Email protected]
Http://blog.csdn.net/kezunhai
This article is based on paper: improving background subtraction using local binary similarity patternsWACV2014 the content and your own understanding, if you want to learn more details, please refer to the original text. The article thought borrowed from vibe, in fact can be understood as VIBE+LBP operator variant (LBSP) motion target detection algorithm combination . In vibe, the algorithm is mainly based on the comparison between pixels, without taking into account the relevant information of domain space. In this article, the author uses the spatial information lbsp descriptors to model the background, improve the detection performance and stability of the algorithm, through the Cdnet data set test shows that the algorithm is superior to most algorithms, the following article is described in detail.
The authors of this algorithm are referred to as lobster (LOcal BinarySimlilarity segmenTER). First, the following LBP operator is introduced, and the formula of LBP operator is as follows.
and lbsp the only difference between this operator and LBP is that the median pixel is the absolute value when compared to the surrounding pixels, or 0 if the absolute is less than a certain value, otherwise 1, the formula is as follows:
Other processing of the algorithm is consistent with the processing of the vibe algorithm. The author also gives the pseudo-code:
The source code can be downloaded here: Https://bitbucket.org/pierre_luc_st_charles/lobster
For more information, please refer to:
1.mproving background subtraction using local binary similarity patternsWACV2014
2. ChangeDetection in feature space using local binary similarity patternsBilodeauetalCRV2013
Kezunhai Source: Http://blog.csdn.net/kezunhai Welcome to reprint or share, but be sure to declare the source of the article.
Moving target detection algorithm based on local two-value similarity mode (LBSP)