Threat mediation via a real-time stereoscopic catadioptric omnivision system

摘要

The authors have invented and are developing a new Real-Time Stereoscopic Catadioptric Omni-Detection (RSCO) system based on a high-resolution stereoscopic vision system, an acoustic array with high-fidelity/sensitivity, real-time image processing hardware based on evolutionary stream processors (SPs) with unprecedented digital signal processing (DSP) performance, and proprietary unwrapping and operating software. The RSCO development involves optimizing (1) the omnidirectional multimodal (visible/MWIR) sensor system configuration and mechanical hardening to survive harsh conditions and to minimize maintenance; (2) a video/image processing hardware architecture based on the ultrahigh-performance SPs, (3) an acoustic array with high-performance DSP technology; and (4) real-time unwrapping and operating software. SPs form a new class of image processors scalable to teraOPS, with efficiency comparable to ASICs, and completely programmable in high-level languages. SPs innovatively combine a new programming model, tool automation, and hardware to exploit the high data parallelism and processing locality that are inherent in a wide range of applications, especially media processing. The excellent performance, efficiency, and programmability of SPs make them ideal for implementation of the RSCO system, with its unprecedented omnidirectional, multiple-modality sensors, range, accuracy, size, and robustness.