"Shape
Activities": General HMM Models for Changing Configurations
of Landmarks and Applications
The
changing
configuration of a group of moving "landmarks" can be modeled as a
moving and deforming shape. The "landmarks" could be moving objects
(people/vehicles/robots) or rigid components of an articulated shape
like the
human body. The term “shape activity” has been used to denote a
particular stochastic
model for shape deformation. We study 3 types of dynamic models:
Stationary
Shape Activities, NonStationary Shape Activities and Piecewise
Stationary Shape
Activities (special case of NSSA). A shape based
dynamical model makes our approach invariant to camera motion, under
the weak
perspective model (also referred to as the scaled orthographic camera)
assumption. The weak perspective model is a valid assumption when the
scene
depth is much smaller compared to distance from the camera and the
centre of
the scene is near the camera's principal axis. This assumption often
holds in
surveillance applications. Also, the
approach is sensor independent, i.e. the observation vector of object
locations
could be obtained by motion detection on a video sequence or using an
infra-red,
radar or acoustic sensor.
Piecewise
Stationary Shape Activities (PSSA) and Applications in Tracking, Change
Detection and Summarization/Indexing of Human Activity Sequences
Talks
UC-Berkeley, Computer Vision Seminar
Series, Nov 22, 2005 (invited): Abnormal "Shape Activity" Detection & Tracking
CDC 2005, Seville, Spain, December
2005: NonStationary
Shape Activities
ICIP 2006, Atlanta GA: Summarizing/
Indexing of Human Activity Sequences
Videos:
Tracking Human
Action Sequences using NSSA:
Samarjit Das's
page
Very Old NSSA videos:
Simulation
sequence: Deforming
Hexagon
Human Actions: Dancer
video (normal),
Tracking and detecting
abnormality
Abnormal Activity Detection using SSA
Group of People
at Airport: Normal activity,
Abnormal activity
Video Summarization
using PSSA (created by Bi Song)
Yoga Sequence
Outdoor
Sequence 1 Outdoor
Sequence
2 Outdoor
Sequence 3 (occlusion)
Indoor Sequence