Abstract
Conventional object recognition systems for robot vision are either fully optical (matched holographic filtering)1 or fully electronic (TV image processing).2 This paper presents a new concept for a hybrid system, which combines the merits of both the optical and the electronic approach but eliminates their respective drawbacks. The resulting system has in its first (optical) part the high speed of parallel Fourier processing1 and in its second (electronic) part the flexibility of digital programming.2 This is realized by replacing the matched holographic filter (hardware) in the Fourier plane by electronic processing (software) of amplitude and phase detected in that same plane. The application of this concept to on-line object tracking and recognition3 will be described.
© 1982 Optical Society of America
PDF ArticleMore Like This
M.S. Monjur, S. Tseng, A.M. Khan, N.M. Rajpoot, and M.S. Shahriar
JTh2A.24 CLEO: QELS_Fundamental Science (CLEO:FS) 2013
B. F. Kim, J. Bohandy, and V. G. Sigillito
FB6 Machine Vision (MV) 1985
Alex Pentland
WA3 Image Understanding and Machine Vision (IUMV) 1989