Background Modeling-Based on temporal and spatial features (Continued 2)

References: spatiotemporal Background Modeling Automation Based on Adaptive Gaussian mixture model 2009

Innovation:

1) Based on the Gaussian mixture model, the spatial distribution information features are mixed.

2). How to make decisions when using the spatial and temporal models for foreground detection.

This article mainly uses the spatial domain background model for rough scale determination,

When the background model of the spatial domain cannot be determined, the background model of the time domain is used for judgment.

3) Adaptive Adjustment of the number of Gaussian Mixture Models.

Time Domain Model:

Added adaptive adjustment module for the number of Gaussian Mixture Models.

For details, see: http://www.cnblogs.com/dwdxdy/archive/2012/05/31/2528790.html

Spatial Domain Model:

Measure the spatial distribution of the background model in the Gaussian mixture model as the spatial background model of pixels.

Color histograms are used to calculate the distribution of background Gaussian components in each pixel's neighborhood.

The formula for calculating the spatial information distribution of the current frame is as follows:

When a pixel matches a Gaussian component of the Gaussian mixture model,

The mean of Gaussian component is used as the statistical sample of the color histogram of the current space,

Otherwise, the current value of the pixel is used as the statistical sample.

Similarity Calculation: histogram intersection. You can use the integral histogram for calculation.

Foreground Detection: