Physical object e.g. aircraft, classifying method for reading e.g. service address on mail, involves parameterizing classifier for classification of physical object with parameter values that are computed during global optimization

摘要

The method involves determining an influencing parameter of an algorithm, which influences a sensor value of a random sample. A preset weighting function is provided as a target function to be locally optimized. A classifier (2) is parameterized with different values for the influencing parameter on a trail basis to compute a local optimum. The parameter is varied using the computed local optimum if the target function is globally optimized. The classifier is parameterized for classification of a physical object with parameter values that are computed during the global optimization. An independent claim is also included for a device for automatically classifying a physical object.