Skip to content

Home

Jump to bottom
Rishabh Kumar edited this page May 16, 2018 · 8 revisions

BackgroundSubtractorMOG is a Gaussian Mixture-based Background/Foreground Segmentation Algorithm Link Here.

WORKFLOW

Run nodes for gmm.cpp, kalman.cpp, publisher.cpp, subscriber.cpp

    rosrun GMM gmm 
    rosrun GMM pub -vid </location of video/> 
    rosrun GMM sub 
    rosrun GMM kalman

GMM

  • Separate foreground from background.
  • Using Contour Detection, detects desired object.
  • Cases are used to differentiate the different detection tasks into.
    • Validation Task
    • Bucket Task
    • Flare Task

PUBLISHER

Publisher takes a video input and publishes it into an sensor_msgs::ImagePtr datatype pointer.

SUBSCRIBER

Subscriber takes the topic input from the GMM node to provide the Radius and Center of the object.

KALMAN

For Template extraction from the detection step, we need to find the object with a certain amount of accuracy. Using KF node we publish the final template to a "measurement" topic which is subscribed by KF-EBT node which is another repository on git.

Clone this wiki locally