system’s components, supporting theoretical ideas that are provided by a Gaussian pyramid, whereasby experimental results. Section 2 introduces a our framework also utilizes intermediate levels. Also,frame-differencingtechnique used to satisfy the real-time execution would require that a high-goal of visually detecting objects of interest. Special performance image processor (like the VLSI chipeffort is made to develop algorithms that allow the developed by van der Wal and Burt'') be available tosystem to function with a consideration for real-time compute the pyramids.constraints. Section 3 addresses additional chal- Theframe-differencingsystembyDinstein''lenges that an automated detection and tracking provides an interesting alternative to the figuresystem must often consider, including analysis of segmentation approach described as a part of ourvisual detection results. Section 4 presents the framework. Dinstein uses four projections to identifyvarious visual calculations that are used in order to the centers of multiple moving objects in a fashionderive the dtsplacements of objects or significant that is only slightly less robust than our method.features within objects. Through these equations, the However, his method locates objects by majority votesystem is able to follow the location of objects as they (much like Hough transforms); as a result, themove. Section 5 describes the hardware modules that method does not provide statistics required byare required to operate the visual detection and other portions of our framework (e.g. it cannottracking framework. Section 6 demonstrates the determinethegeometriccharacteristicsof anframework’s use in robotic visual serx oing. Addi- identified object).tional issues for the integration of a detection module Considering application-specific detection, Allen et
