| 摘要 |
The invention concerns a method for recognising the forms of objects defined by a digital object image (IO) which consists in defining, in a prior step and independently of each object image, in a reference region, an array of intersecting subspaces (gj). For each object image, computing by projection of the reference region (RR) on the object image a circumscribing region (RC) having a size corresponding to that of the object image (IO); defining an array of applied subspaces (gj'') by affine application of each intersecting subspace (gj) to the circumscribing region (RC); computing and storing the number of intersections (qj) which each applied subspace (gj'') has with the object image (IO) and using said number of intersections (qj) to recognise the object form, by comparison with those of known object forms. |
