Students of IIT-Delhi have developed a system that could give cops a shot in the arm, reports Meenal Dubey.india Updated: Apr 20, 2006 14:03 IST
Students at IIT-Delhi have developed a surveillance system which could give police a shot in the arm.
What's more, the system — after customisation — can come in handy at public places like malls, airports, railways stations, parking lots and Metro stations.
The gadget, which will be on display at the forthcoming 'Open House', is portable and can be customised for any scenario with minimal modifications.
The system has been developed by fourth-year students of Computer Science and Engineering — Nitin Jindal and Shubham Singhal.
"We have developed a fully automated surveillance system that generates intelligent object information from surveillance videos in realtime," said professors Prem Kalra and S Bannerjee who supervised over the project.
"This information can be used for activity recognition in both supervised and unsupervised manner. The system is divided in three layers."
The lowest layer, which is a crude layer, detects objects in the scene. "We learn the background of the scenario and then use background subtraction for any new object entering the scene.
"The second layer, physical layer, generates preliminary intelligent information about the object in this layer. We recognise the posture (standing, sitting, bending). We then fit a 6-ellipse human body model, two hands, two legs, head, torso of the object. We also recognise the human being based on their faces by creating high resolution images of face by super resolution technique," explained Shubham.
"The third layer, logical layer, incorporates the temporal information of the scene. We handle occlusion of multiple objects based on the colour model learnt for the objects. We are also able to locate the accurate ground position of the object in the scene, hence giving us his depth information," said Shubham.
The last layer, event layer, is used to recognise the activities. It uses the intelligent information generated by the three layers to recognise pre-known activities.
This layer can be customized according to the requirements of the scenario like detecting if a person has left a bag in some place, or for shop lifting detection.
The crux of this layer is to use the above generated information to suit the activity recognition requirement.
The students have also developed an ultra violet protection umbrella which provides screening from harmful solar radiations. It has a high value of ultra violet protection factor.
Other interesting projects include "defluoridation" apparatus of underground water. Fluoride is an acute toxin.
Its contamination is endemic in 19 states and nearly 65 million peoples, including six million children, are affected.