CrowdPic: A Multi-coverage Picture Collection Framework for Mobile Crowd Photographing

This paper proposes a generic task-driven data collection framework, named as Crowd Pic, for Mobile Crowd Photographing (MCP) - a widely used technique in crowd sensing. In order to meet diverse MCP application requirements (e.g. Spatio-temporal contexts, single or multiple shooting angles to a sensing target), a multifaceted task model with collection constraints is provided in Crowd Pic. Meanwhile, a pre-selection process is necessary to prevent mobile clients from uploading redundant pictures so as to reduce the overhead traffic and maintain the sensing quality. To address this issue, we developed a pyramid-tree (PTree) model which can select maximum diversified subset from the evolving picture streams based on multiple coverage requirements and constraints defined in MCP tasks by data requesters. Crowd sourcing-based and simulation-based methods are both used to evaluate the effectiveness, efficiency and flexibility of the proposed framework. The experimental results indicate that the PTree method can efficiently assess redundant pictures and effectively select minimal subset with high coverage from the streaming picture according to various coverage needs, and the whole framework is applicable to a wide range of use scenarios.