Efficient photo crowdsourcing with evolving POIs under delay-tolerant network environment

Abstract

In a disaster or battlefield zone, rescue workers, soldiers, and other survivors (referred to as nodes) may need to survey damages and send images to the command and control center (the server) in a hop-by-hop fashion in the absence of any communication infrastructure. The server considers some area/landmark as the point of interest (POI) and distributes a request to the nodes to collect more information about them. Nodes take photos of POIs and share them with each other using the store and forward paradigm, also called Delay-tolerant Networks (DTNs), to send the photos to the server. Due to the highly intermittent contact characteristics of nodes in a DTN network and bandwidth and storage limitations, redundant photos need to be removed in this forwarding technique, whereas photos that cover different angles and views of the POIs need to be shared. Another challenge is that, over time, some server-listed POIs may not be of importance anymore, whereas some new POIs might be of more interest. In this work, we propose a scheme that is able to dynamically update the list of POIs based on the current photo metadata, with reduced consumption of the bandwidth, energy, and storage at DTN nodes by sending only important photos of POIs. In the performance evaluation, we show the scalability of our approach, and also show that it provides the same level of photo coverage but consumes much less energy and bandwidth than other related schemes.