Download CitationEmailPrintRequest Permissions

An improved algorithm for circular parts detection was proposed based on RHT (randomized Hough transformation). At first, sub-pixel edge detection based on the Zernike moments was chosen to detect edge and get the more accurate target edge positioning. Then, a regular hexagon window, whose center was the candidate circle's center, was selected as search constraints to reduce the computation time and increase the calculation speed. At last, the marginal gradient direction information was used as constraint conditions to limit retrieval object. Thus invalid computation was reduced and the detection speed was accelerated. The proposed algorithm was used to detect circular parts size. Experimental results show that the measurement error is less than 4.5%. And the performances of the non-contact inspection system can meet the demands of on-line and real time measurement with high accuracy in the production line.