Automated Target Detection from Compressive Measurements

摘要

A novel compressive imaging model is proposed that multiplexes segments of the field of view onto an infrared focal plane array. Similar to compound imaging, our model is based on combining pixels from a surface comprising of different parts of the FOV. We formalize this superposition of pixels in a global multiplexing process reducing the number of detectors required of the FPA. We then apply automated target detection algorithms directed on the measurements of this model in a scene. Based on quadratic correlation filters, we extend the target training and detection processes directly using these encoded measurements. Preliminary results are promising. This work is based upon work supported by DARPA, Air Force Research Laboratory / RWWI, and the SPAWAR System Center Pacific under Contract No. N66001-11-C-4092.