| 摘要 |
The invention relates to a method for assessing similarities of real objects based on an assessment of the similarities of feature vectors. According to said method, sets of one-dimensional parameterized data (feature vectors), which constitute the basis for the comparative operation, are obtained from the 2D contours of the first and second objects that are to be compared to each other. The similarity of the feature vectors is calculated taking into account the phase similarity of the individual vector components on the complex plane. The similarities are calculated separately according to X features and Y features for each complex coefficient of the feature vector, whereupon the results are multiplied such that a great similarity is obtained only if both the similarity of the respective complex X coefficients and the similarity of the complex Y coefficients are great. The inventive method allows data that is provided only in part or is extended by additional information or interferences to be recognized in a translation-invariant and scaling-invariant manner, making the recognition comparable with human perception. |
