Learning Integrated Recognition for Image Exploitation
Bir Bhanu · Defense Technical Information Center (DTIC) · 2003
The overall goals of the proposed learning integrated object recognition for image exploitation research effort at the Center for Research in Intelligent Systems of the University of California, Riverside are to improve the performance and reliability of automated systems that can recognize objects in reconnaissance imagery acquired under dynamically changing conditions and for systems that can efficiently extract information from enormous image databases. This requires innovative techniques developed through fundamental scientific research in the fields of machine learning and computer vision. The research accomplished in this effort involves four specific areas: (1) Predicting the performance for recognition systems; (2) Automating methods to develop composite class models for SAR recognition; (3) Learning integrated physics-based fusion of IR and video for target detection; and (4) Learning concepts in images/videos. This report summarizes the achievements in each of the four major research areas.