Background Subtraction and modeling common Background Modeling Methods in Target Detection

For the interpreter:

Recently, I have been reading background subtraction and modeling. As the first reply in the following link says, although the recent research on Background Modeling is not very popular, it is crucial for video processing and even directly affects the success or failure of a system. The reason for the study being not hot may be that the methods that can be mined are almost the same, and the rest is personal digestion and understanding. Sometimes a method may not work well. It may take several combinations to make it effective, and different parameters are required in different scenarios, which can be determined in a specific experiment. In this sense, it is meaningful to understand the existing algorithms for background subtraction and modeling. You can extract the desired methods based on your specific tasks and requirements. No more nonsense.

= ======================================

Http://cvchina.net/thread-2351-1-1.html

I have been doing research on foreground detection recently. At the beginning, I mainly made some engineering applications. I have made a lot of effort to solve engineering problems, I have also read many recent articles at home and abroad. I have always wanted to make a conclusion, dragging on, but after all, I wrote this extremely unsuccessful summary. Background Modeling or foreground detection algorithms mainly include:

1. single Gaussian (single Gaussian Model)

Real-time tracking of the human body

2. mixture of Gaussian Model)

An Improved Adaptive Background Mixture Model for real-time tracking with Shadow Detection

3. Running Gaussian average --- single Gaussian

Real-time tracking of the human body

4. codebook)

Real-time foreground-Background Segmentation Using codebook Model

Real-time foreground-Background segmentation using a modified codebook Model

5. Self-Organizing background detection (Sobs-self-organization background subtraction)

A self-organizing approach to background subtraction for + visual surveillance

6. Sample consistency Background Modeling Algorithm (sacon)

A consensus-based method for tracking

A consensus-based method for tracking-modelling background scenario and foreground appearance

Sacon-background subtraction based on a robust consensus method

7. Vibe Algorithm

Vibe

Vibe-A Universal Background Subtraction

8. color-based Background Modeling (color)

A statistical approach for real-time robust background subtraction and shadow detection

9. Statistical Average Method

10. Temporal Median Filter)

Automatic congestion detection system for underground Platform

Detecting Moving Objects, Ghost, and shadows in Video Streams

11. W4 Method

W4.pdf

12. Intrinsic background

A Bayesian computer vision system for modeling human interactions

13. Kernel Density Estimation Method

Non-parametric model for background subtraction

We are familiar with single-GAOs and mixed Gaussian estimation. We will not describe it here (the mixed Gaussian model should be better in the existing background modeling algorithms, many new algorithms or improved algorithms are based on different variants of some of their principles. However, the disadvantage of the Gaussian mixture algorithm is that the computing workload is relatively large, the speed is slow, and the algorithm is sensitive to light ); I have performed experiments with codebook algorithms, and the results are still good. There are also many variants later. I have not studied them further, but the algorithms are also sensitive to illumination ); for the self-organizing Background Modeling Algorithm (sobs), this algorithm is robust to illumination. However, the map model is larger than the input image, and the computing workload is relatively large, however, you can solve the speed problem of the algorithm through parallel processing. The sacon algorithm is based on statistics knowledge, has been implemented by code, and has been tested. The effect is also acceptable, but there is no further analysis. The vibe algorithm is a masterpiece of B's brother. There are ready-made algorithms available on the Internet, but it has applied for a patent and can be used for research. The algorithm is very fast, the algorithm has a relatively small amount of computing and is robust to noise, and the detection effect is good. The Background Modeling Method Based on Color information, referred to as the color algorithm Pixel differences are divided into chromaticity differences and brightness differences, which are robust to illumination and have good results. The computation speed is also fast and can basically meet real-time requirements, we have done a lot of video sequence detection, and the results are quite satisfactory. The statistical mean method and the median filter method are only implemented for these two algorithms and tested, the Application of algorithms has great limitations and can only be regarded as a supplement in theory. The W4 algorithm should be the first algorithm used for practical application. You can check the relevant information, I will not elaborate on it here; the intrinsic background method has not been implemented. I have read a lot of literature to explain it, and then the algorithm is based on the Bayesian framework. I have always been familiar with the Bayesian framework. In theory, perfect is very good, the actual application is shit (if you do not want to offend the fans of beishi, Please bypass the road and do not vomit). Finally, it is the kernel density estimation algorithm, which should be a relatively robust algorithm, it can solve many algorithm parameter settings problems, and does not need to set parameters should be a major advantage of the algorithm.

Personal Opinion: sobs, color, vibe, sacon, PDF, etc, in particular, the block-based or region-based, features-based, hierarchical classification or hierarchical trainer algorithms that have emerged in recent years can be thoroughly studied.

We recommend a summary article: Evaluation of Background Subtraction Techniques for video surveillance

Recommend another website: http://www.changedetection.net/