Class ConfigBackgroundGmm

All Implemented Interfaces:
Configuration, Serializable

public class ConfigBackgroundGmm extends BaseConfigBackground
Configuration for ConfigBackgroundGmm.

Tuning Advice: Background models do a much better job of deciding if pixel doesn't match a model it has. This background model has multiple Gaussians. So you should keep the number of Gaussians small so that when the moving object goes past it doesn't because part of the model right away. If you have too many Gaussians its basically impossible to keep the amount of noise down and have a stable foreground object.

See Also:
  • Field Details

    • learningPeriod

      public float learningPeriod
      Specifies how fast it will adjust to changes in the image. Must be greater than zero.
    • initialVariance

      public float initialVariance
      The initial variance assigned to a new pixel. Larger values to reduce false positives due to under sampling. Don't set to zero since that can cause divided by zero errors.
    • decayCoefient

      public float decayCoefient
      Determines how quickly a model is forgotten. Smaller values means they last longer. Values equal to or greater than one might cause new Gaussians to be immediately removed.
    • maxDistance

      public float maxDistance
      Maximum Mahalanobis a value can be from a Gaussian to be considered a member of the gaussian
    • numberOfGaussian

      public int numberOfGaussian
      Maximum number of gaussians that can be in a single mixture
    • significantWeight

      public float significantWeight
      Once the weight for a Gaussian becomes greater than this amount it is no longer considered part of the foreground and is the the background model. Strongly influences how long it takes an object that was moving to fade into the background. Probably one of the first tuning variables you should mess with.
  • Constructor Details

    • ConfigBackgroundGmm

      public ConfigBackgroundGmm()
  • Method Details