Abstract: In order to improve the detecting efficiency of the crankshaft shoulder,a machine vision system for crankshaft rapid detecting was developed.Firstly,according to the crankshaft grinding process and the producing principle of over limit error,the image identifying criterion to judge the acceptable crankshaft part and the nonconforming product was provided.Secondly,the detection platform of the crankshaft with part-hold and light imaging system was designed.Some high-efficiency image processing algorithms was described.Finally,the vision system performance testing experiments were carried out.Experimental results indicate that the detecting accuracy rate of the crankshaft vision detection system is 100% and the detection system is easy to use.The detecting system has high measurement efficiency,which could be used to check many shaft parts in manufacture.

Key words: computer vision, parts detection, crank shaft shoulder, image processing