Areas of Interest

Thesis work

Crowd Counting and Tracking in video Sequences: Monitoring crowded urban environments is a vital goal for many of today's vision systems. Knowing the size of crowds and tracking their motion has many applications. For example at traffic intersections, intelligent walk-signal systems could be designed based on the number of people waiting to cross. Also, the knowledge of the number of people walking through a crowded area, e.g., outside a school or outside the premises of a public event can be helpful in planning urban environments, general safety, and crowd control. We estimate accurately the counts of people in a scene without constraining ourselves to individuals. This includes dense groups of people moving together. We do this in real-time and place no constraints as far as camera placement or about the size of the groups as far as number of people.

SAFAR: An Adaptive Bandwidth-efficient Routing Protocol for Mobile Ad hoc Networks: A mobile ad hoc network suffers from the same cost constraints as most wireless networks. In particular bandwidth constraints of wireless links are severe. We present a scalable adaptive fitness-based routing protocol, SAFAR, for mobile ad hoc networks in which we try to optimize the usage of this bandwidth at every stage. The protocol is hybrid, i.e. it makes use of both proactive and reactive procedures for routing in an attempt to reduce route acquisition latency. Using a fitness function a node decides how many other nodes can be proactively maintained by it. Each node tries to know the best nodes in its neighborhood. Hence, high bandwidth nodes are well known. Most of the traffic is routed through these nodes and hence performance is optimized. We present simulation results to substantiate the protocols performance. We also extend this protocol to show how it can be used for power aware routing.