| OCR Text |
Show 77 irnplenlel~ted and tested based on these two approaches. This shows that combining the adaptive-single-tolerance approach and redundancy avoidance is a practical and efficient way of solving the robustness problem in geometric modeling. The ideas in this thesis can be applied to a wide area where symbolic information has to be constructed frorn approxirnated nun1erical data. There are sirnilar problen1s in pattern recognition ( cornputer vision) and data reduction as used for instance in scientific visualization. In order to recognize an object frorn noisy in1age data, a dynarnic tolerance can be used as a feedback reducing the noise and reaching a consistent interpretation of the object. Sin1ilarly, by associating a dynarnic tolerance with sampled noisy data, the consistency in the interpretation of n1easured data by mathernatical relation in scientific visualization can be ensured. Tolerances are irnportant in rnechanical design, manufacturing, inspection, and reverse engineering. The advantages and irnplications of a tolerance-based rnodeling approach for these areas are currently investigated. |
