Over the last few years, I have been working on the following projects. For code and datasets follow the links below.
Human Activity Recognition and Tracking
Developing methods for simultaneous tracking and recognition of
humans performing various activities in a scene. This involves posing
the two problems in a unifying optimization framework that uses the
tracked features to perform better recognition and then uses the
recognized activity to perform better tracking.
Developing scale and direction invariant Histograms of Oriented Optical Flow (HOOF) features to represent action profiles at a time instant. The non-Euclidean time-series thus generated from the activity video can be modeled as a Non-Linear Dynamical Systems (NLDS).
Developing an algebraic method for classification of non-Euclidean time-series data based on the Binet-Cauchy kernels for NLDS using Mercer kernels defined on the non-Euclidean space.
We provide code for computing Histograms of Oriented Optical Flow (HOOF) - which are features computed at each time instant from a frame of optical-flow vectors to model dynamic phenomena such as human activities. Once a time-series of HOOF is extracted from a video sequence, these can be modeled as Non-Linear Dynamical Systems (NLDS). Kernel PCA is used to learn the parameters of these NLDS and the Binet-Cauchy kernels for NLDS can be used to compute a distance between pairs of such NLDS. Dowload (registration required)
Classification of Linear Dynamical Systems
Modeling the following visual processes as linear
Using System ID techniques to recover the models
Classification, Clustering and Recognition of the
The dynamic texture toolbox contains implementation of methods for identifying and comparing video sequences by representing them as Linear Dynamical Systems (LDSs). All code is currently implemented in MATLAB with some code depending on the MATLAB control systems toolbox. A set of tutorials on how to use the toolbox can be found here and the complete documentation can be found here. The code is licensed under the Lesser GPL and is provided for academic purposes. In case of questions, comments, bugs or errors, please contact Avinash Ravichandran (avinash at cis dot jhu dot edu) or Rizwan Chaudhry (rizwanch at cis dot jhu dot edu). Dowload (registration required)
Visual Sensor Networks
Using a network of low power wide-angle lens cameras on Intel
Imote2 processors to reconstruct the 3-D structure of a scene
Vision based control of Unmanned Autonomous Vehicles (UAV)
Developing vision based algorithms to control the
path of a UAV and guiding it through a terrain with potential