Image Feature Extraction for Improved EW Classification

The proliferation of radar and communication devices significantly increases operators"workload. Different platforms carrying similar communication devices further complicate the problem. The onboard imaging sensors, combining with EW sensors, have the ability to disambiguate devices and platforms. Our proposed system consists of three main modules: (1) the data/time synchronization and sharing module, (2) the image processing and feature extraction module, and (3) the Bayesian data fusion/classification module. Our proposed system integrates the tracking function with region of interest segmentation and feature extraction. We propose the use of prior probabilities assigned to feature elements to handle the problem of data synchronization, feature quality and object occlusion in a unified manner. We fuse the prior class probability distribution estimated from previous instances of the same targets with the new classifier output to establish the posterior class probability distribution in a Bayesian framework. The project team consists of Scientific Systems Company, Inc. (SSCI) as the prime contractor and Georgia Institute of Technology as the academic partner.