Robust Data Alignment and its Application in an Integrated Multi-Sensor Surveillance System Consisting of UAVs and Fixed Platforms. The Current and Future Phase of the RASER Project

摘要

This report introduces a new algorithm called Robust Data Alignment (RDA) for persistent sensing in multi-layered sensor network. Invariant features collected from a set of images are used for image registration. Information-theoretic cost models a functional structure where an optimal transformation can be recovered from feature representations of different but related images. By using a cooperative search strategy, we believe that RDA can help to understand and deal with many situations in persistent sensor networks by solving problems such as Layered sensing, Multi-modal data fusion, Multi-UAV sensing. In addition, we provide Georegistration for CLIF 2007 dataset. SIFT algorithm after georeferencing recovers a correspondence for data registration.