A system for shape recognition of simple objects (a cube, a plate, a rod) falling in the field of vision by extracting their image contours is proposed. Recognition is carried out from three projections. The minimal time of image processing is 1 ms. The idea of the simultaneous extraction of an image contour and the parameters characterizing its shape is proposed based on the properties of human vision. The sensor is an optoelectronic functional converter representing a system for searching and extracting an image contour by the radius-vector scan with a dissector as a photodetector. It realizes the fastest method of object image analysis to recognize image shape because of the linear functional connection between an electronic parameter (a brightness level as an argument) and a geometric parameter (for example, a radius as a function) of an image in the process of following its contour with a photodetector aperture. As a result, the instantaneous video signal amplitude (which is proportional to the brightness level at the point of the aperture location) corresponds to the current length of the image radius, and its variation during the tracing of the contour adequately characterizes the image shape, including such features as the extreme dimensions and the angles. Simultaneously the coordinates of the contour points of the image and its center of gravity are extracted. The highly informative nature of the video signal also enables us to determine the perimeter, the area, the orientation of the image, and the coordinates of its brightness extremum. The radius-vector scan is the only sensor currently available whose signals carry direct information concerning the image shape in real time. # 1998 Society of Photo-Optical Instrumentation Engineers, S0091-3286(98)00310-9 Subject terms: shape recognition; vision system; contour following; radius-vector scan. Paper 29067 received July 3, 1997; revised manuscript received Apr. 16, 1998; accepted for publication Apr. 22, 1998.
展开▼
