Incorporating Scene Mosaics as Visual Indexes into UAV Video Imagery Databases

Abstract : The rise of large digital video archives has strengthened the need for more efficient ways of indexing video files an accessing the information contained in them. Reconnaissance platforms, such as the Predator UAV, are contributing thousands of hours of video footage that require analysis, storage, and retrieval. A process is proposed for converting a video stream into a series of mosaic and selected still images that provide complete coverage of the original video. The video mosaic images can be utilized as visual indexes into a video database. in addition, mosaic images contain information from an entire sequence of video frames to provide "at a glance" analysis capabilities. Actual reconnaissance video footage is converted to still-image representation using the proposed process and the results are discussed. Further, a web-based browse and search capability was developed to demonstrate the benefits of using the proposed process. Further, the Predator Unmanned Aerial Vehicle (UAV) system configuration is described with recommendations for placement of the video mosaic building process proposed in this research.