Professional Documents
Culture Documents
Name
ID No.
CGPA
Contact No.
Karan Chawla
2012A4PS338G
8.3
9637107801
2012A8PS321G
5.9
3.
Siddharth Jadhav
2012A1PS509G
7.1
karanchawla.ck@g
mail.com
akash.singh011235
@gmail.com
f2012509@goa.bi
ts-pilani.ac.in
9628963929
9021217959
PROJECT
PROJECT INFORMATION
INFORMATION
1. Title of project
One of the major aim of our project is to show how a visual human detection
system that uses images from a normal camera can be implemented in
software. The system would fit into the larger context of disaster
management and more specifically how it can benefit search and rescue
operations. The image processing and computer vision capabilities reduces
the psychological strain on the human operator.
The final design would be able explore an area entirely by itself and human
intervention would only be needed if a victim was discovered. Simultaneous
Localization and Mapping algorithms are implemented that allow the UAV to
map an unknown environment by using a variety of sensors. The mapping
information is combined with the detection system by recording the position
and orientation of the drone when a victim is detected. This would aid rescue
workers in locating and evacuating victims.
The necessary research was then done to solidify the concept post which the
development of the airframe was initiated. The design was then documented
and the first prototype is constructed post which the next set of iterations will
take place. The first prototype of the design was the Discovery UAV which
was used to collect data in the form of videos and the autonomy of the
system was successfully tested.
Image 3:The result of the HOG detector tested on a normal camera feed
A pair of temporarily and geometrically synchronized optical images are used for
human detection. First these images are processed independently and then related
information from the image sources is combined. For detection the optical images
are processed with several independent cascaded Haar classifiers trained for
detecting humans in different orientations and light/dark colors. Next the
Implement and evaluate algorithms for object detection based on a live feed
from a camera.
Implement data logging of the tracking results, video streaming and a simple
vision-based navigation algorithm.
Make the drone portable. Further optimize the design to increase range and
endurance of the plane.
5. Which stream would you identify your project most closely with? (Comp
application.
Rapid prototyping is the future of drones since it is highly cost effective and
will help the user deploy these from any location. The final aim would be to
create a 3D printed version of the drone that could be produced in a matter
of hours. The electronics for such drones would be designed in the form of
LEGO blocks which would reduce assembly time to minutes.
7.