Straight-line segments are often used as key features in analyzing scenes. In this paper, we extract straight-line segments using advanced Paton's template masks (1979) for straight-line segments detection in real time. To overcome a problem of a complexity of scenes and simplify straight-line segments, straight-line segments are grouped into longer straight-line segments by collinearity of perceptual organization. After the existence of undetected-line segments by parallelism and symmetry of perceptual organization is supposed, undetected-line segments are extracted by applying a new threshold again. Here, feedback control can be seen. Data are integrated by bottom-up and top-down method. To control thresholds in a structural manner, a fuzzy logic controller is used. By extracting straight-line segments of natural scenes and man-made objects, the effectiveness and the advantage of this method are shown. This proposed approach can be applied to a pattern recognition.
展开▼