Background Subtraction (Foreground-Background Segmentation) with Adaptive (online) GMM, Tracking with Pixel Process on an input traffic video (from mathworks) - an implementation with python
- A time series of feature values (pixel process) is approximated from the recent frames with k (= 3 here) Gaussian mixture models (per-pixel)
- The parameters of the GMM updated online using the Stauffer-Grimson algorithm
- The contour plot of the GMM for a pixel (e.g., the pixel at location (25, 460)) is visualized, with the alpha value (transparency) of a contour is proportional to the weight of corresponding Gaussian component
- The weight of a Gaussian from the mixture components increase when there is more evidence in favor of it
#imageprocessing #imageprocessingpython #python #computervision #machinelearning #algorithm #gaussian #tracking #videoprocess
