| 摘要 |
PROBLEM TO BE SOLVED: To surely specify the necessary number of feature points by obtaining a zero cross image by means of a DOG(difference of two Gaussian) function in execution of the pattern matching and then producing a binary image of a zero cross point to perform the approximation of a polygonal line. SOLUTION: The image processing is performed by a DOG function and a black/white binary image using a zero cross point as a boundary is obtained from a zero cross image (edge image), for example. Then an entire white area peripheral coordinate is extracted out of the binary image and the approximation of a polygonal line is performed in consideration of the error of a ±1 pixel. An end point of the polygonal line decided by the said approximation is defined as a feature point. As a result, a general graphic form can be expressed as a feature point even if an object has a small number of clear feature points and also can be recognized by means of the DOG function despite the change of brightness, the influence of a background and the change of the object. |
