Verfahren zur automatischen Erkennung von auswertbaren Zonen in Bildern von mechanischen Teilen

摘要

The image detection method involves segmenting reference images stored in a learning base manually, one at a time during a first phase (1). An operator indicates the number of zones and marks them with a pointer and the exact contours of the zones are then determined using the method of parting waters. The second phase (2) consists of an optimisation of the marking parameters and the calculation of zone statistics and grey level histograms. The third phase (3) uses the marking parameters obtained in the second phase to produce mark the zones on new images completely automatically.