of image understanding architecture: studies in automatic target recognition system design

Three generations of image understanding architecture: studies in automatictarget recognition system designWilliam W. Wehnerr Paul S. SchenkerSignal and Image Processing Groupf Systems and Research Center, Honeywell Inc. MN17-2306, 2600 Ridgway Parkway, Minneapolis, Minnesota 55413AbstractThe increasingly broad scope of intended image understanding (IU) applications is driving IU architectures toward general purpose designs. This is reflected in growth of first generation domain-and-function specific processors to new multi-function/multi-scenario designs. These advances have been enabled by novel algorithm developments, expansion to multi-sensor capability, advances in VLSI/VHSIC circuit technologies, and development of supporting software design methodology.Our paper presents a perspective on on-going and anticipated developments in military image understanding system architectures. We briefly discuss the types of missions and applications motivating system developments. We overview resulting system requirements and classes of supporting algorithms. We discuss resulting processor requirements and show by case study how we are addressing them in our past and present IU system designs. The trend we establish is development from early applications-specific hardwired processors to future generation modular, reconfigurable, high-level programmable VLSI system architectures.IntroductionA variety of technologies for image understanding (IU) applications and algorithms, digital electronics, computer architectures, computer software, and system development methodology are driving architectures for image processing and image understanding. The complexity of the IU problem together with the push-pull relationship between its support technologies tends to obscure both the high-level requirements driving image understanding architectures and the resulting system trends. Our work in the development of IU systems and support technologies has forced us to face these system considerations. In this paper, we examine IU system requirements, resulting processor requirements, and processor design methodologies. We focus particularly on our experience with the automatic target recognition (ATR) problem and we attempt to separate the application dependent and independent features of architectures we have used in ATR systems. Finally, we present a brief study of several IU architectures for automatic target recognition defined under Honeywell sponsored research and currently being realized in on-going programs.System requirementsFigure 1 shows a few of the suggested and planned military systems in which innovative image understanding (IU) architectures are finding or may find application. Each illustrated system is designed to meet specific mission goals. To attain these goals, the system designer imposes constraints and requirements on IU processors and other subsystems. Some of the requirements are forced on the IU processor (e.g., the operational temperature range), but other requirements may be strongly influenced by the processor designer (e.g., type of sensor or sensors required) provided a sufficient understanding of the system and mission environment exists. The goals of the processor designer are to develop IU subsystems which meet the system requirements and take advantage of developments in IU technology to help shape future system directions.General system requirementsStringent military requirements are forced on the IU processor architect working on the applications suggested by Figure 1. The IU subsystem must meet severe space, weight, and power constraints. These constraints are further complicated by operating environment and system survivability, reliability, and maintainability requirements. The designer must also be sensitive to the life-cycle cost of the subsystem as a function of unit cost, maintenance cost, total units produced, and upgradability of units as missions and systems change. Although these are the most obvious requirements imposed on the