HumanPosedetection人体姿态检测

*,Click to edit Master title style,Click to edit Master text styles,Second level,Third level,Fourth level,Fifth level,Human Pose detection,Abhinav Golas,S. Arun Nair,Overview,Problem,Previous solutions,Solution, details,Problem,Segmentation of humans from video capture,Pose detection (by fitting onto body model),Resistant to noise (background etc.),Previous procedures,View problem as sequential process,Segmentation,Pose detection,Problems:,Not using prior knowledge of “what a human looks like in segmentation,Uses only information from detected “foreground for pose detection,All available information not used,Solution,Combine segmentation and pose detection as a single step,Uses all available information in frame (for pose detection),Uses prior knowledge of human body for better segmentation,PoseCut: Bray, Kohli, Torr,Model segmentation as Bayesian labeling problem with 2 labels: foreground, background,Details,Model problem as energy minimization problem model as an MRF,Use a basic stickman model as a human body model,Adaptive model for background GMM,Neighbourhood terms Generalised Potts model,MRF Markov Random Fields,Markov property for time:P(event:t) depends on events at times kt,Markov property for space:P(event:x) depends on events at N(x) neighbourhood of x,Use Gibbs energy model for solving,We use neighbourhood of 8 pixels,Stickman model,Basic model,26 degrees of freedom,GMM Gaussian Mixture Model,Model each pixel of image as a weighted sum of Gaussian functions,Adapt functions using each new frame,Pixel matches expected value background, else foreground,Execution details,For each frame,Calculate weights for GMM, Potts model,For given value of 26 vector (based on degrees of freedom of stickman model) calculate energy cost for stickman model (by distance transform),Minimize energy for Bayesian labeling by graph cut,Minimize 26 vector by repeated graph cuts by Powells algorithm,Sample results,A original frame,B segmentation by colour likelihood and contrast terms,C when GMM terms are taken,D with pose prior components,E deduced pose,Comparisons,