Vision Based Online Human Tracking Using Dynamic Object Model

There are
countless objects present in our surrounding environment with their impressions. Vision can be
better explained as a way to understand the environment that surrounds us. Even after decades,
the exact working of the visual system is a mystery for the scientist involved in its
investigation. When eye based vision in living creatures is replaced by computational instruments
it is defined as computer vision. In other words, computer vision is artificial mimicry
of vision in living creatures, where digital images and videos which are captured by
cameras are further analyzed by computers obtaining an optimum level of understanding from it.
Human/object tracking when done on an array of frames is an operation of tracking any mobile
target object over a span of time with the help of any mobile or immobile camera. It
has been a critical issue in the arena of computer vision as it is used in a
number of application fields like security, surveillance, human-computer interaction,
augmented reality, video communication, and compression, medical imaging, traffic control,
video editing, and assistive robotics . This is a highly studied problem and remains to be a
complex problem to solve. In object tracking in any given video, the
major task is to trace the target object in upcoming video frames.

Object tracking is a principal segment of human-computer interaction in a real-time
environment, where the computer obtains a finer model of a real-time world. For
example, when autonomous vehicles are talked about, a human being cannot transmit the exact state
of surroundings precisely and speedily enough.
The wide-ranging scope of the application review the significance of dependable, exact, and
efficacious object tracking. To obtain an effective tracking the two most important parameters to
be included are first, selection of the model and secondly, the tracking
method worthy for the task.

The fundamental necessities of any tracking structure are first, a robust system,
secondly, an adaptive system, and lastly, real-time processing requirement . The
famous state-of-art tracking strategies are Interest point-based tracking , multiple
hypothesis tracking , kernel-based tracking , and optical flow-based tracking . This
area has observed a remarkable elevation due to available low cost, advanced technology cameras,
and low computing complexity, corresponding to the inclination of ingenious approaches for image
and video processing. Excellent reviews on the state-of-
art techniques in this area have been provided in