Fusing geo-referenced images for urban scene

Advances in surveillance technology create challenges for management and presentation of massive amount of sensor and imagery data. The imageries acquired by the surveillance sensors often yield “soda straw” views of the situation due to the camera narrow field of view (FOV). Conventional surveillance systems that comprise of multiple screens, each displaying respective acquired video stream, fail to convey spatial relationships among the views. To provide enhanced situational picture of wide area scenario, a promising solution involves fusing and visualizing multi-view imageries spatially in common three-dimensional (3D) space. The capacity to visualize and simulate in a virtual environment with synthetic urban terrain modeled using the latest acquired imageries in real time allows novel insights to be gained. Fusing multi-view imageries spatially in complex urban environment, however, is computationally expensive and typically non interactive. This paper presents a proposed approach to visualize and fuse spatial imageries acquired by different surveillance devices planted in urban environment. In light of availability of geometrical information due to recent advances in 3D acquisition and reconstruction technologies, the proposed approach utilizes an integrated environment that considers inter-object occlusion and fuses relevant parts extracted from imageries. The concept focuses on harnessing graphics rendering techniques and exploiting Graphics Processing Unit (GPU) programmability. The approach fuses geo-referenced imageries with synthetic terrain and assumes that acquired images are coupled with corresponding camera parameters. The assumption is coherent with the advent of ubiquitous Global Positioning System (GPS)-enabled devices and related technologies.