Research Of Detection Methods Based On Vision Apply For The Quality Of O-Ring

O-ring is a kind of potted component in the industry, which is the most widely applied and has the simplest structure. O-ring is used extensively in the petrochemical engineering, the shipbuilding, the Aeronautics, the Astronautics and the military hardware. So it is very important to ensure O-ring of the high quality. At present, the detection of O-ring’s quality mainly rely on manual detection in our country. When facing mass inspection,it will be not only time consuming,low precision and large Labor intensity but also influenced by the subjectivity of the quality inspectors.The paper mainly research the detection algorithms based on vision apply for the quality of O-ring, which include the accurate measure algorithm of the internal diameter and sectional diameter and the identification and classification algorithm of the surface defects. At last, the paper build a vision based O-ring quality detection system.First of all, anti-noise multi-structuring elements morphological edge detection algorithm is proposed. Combining the characteristic of the basic algorithm of mathematical morphology and different structure elements, the paper improves the traditional morphological edge algorithm and acquire the algorithm. The algorithm can effectively eliminate the reflective edge of O-ring surface and the local small defect edge.Second, improved method of least squares round segmentation fitting algorithm is proposed which is based on subpixel edge. Introducing round features estimation equation improves least squares method of fitting round, eliminates the impact of small errors estimate and fit the segmented inside and outside round of O-ring on the basis of the subpixel edge. When measuring the internal diameter and the section diameter, it reduces the measurement error of O-ring.Then, complete kernel fisher discriminant analysis algorithm is proposed. In the defect inspection and defect classification algorithm, the paper creatively uses pattern recognition algorithms, introduces a second type of authentication information on the basis of KFD algorithm, form a complete fisher discriminant analysis algorithm(CKFD). The experimental results show the CKFD algorithm is much higher than KPCA and KFD on defect recognition rate and classification accuracy.The last, the O-ring visual on-line inspection system is developed. Through trials, it further validates the applicability of anti-noise multi-structuring elements morphological edge detection algorithm, improved method of least squares round segmentation fitting algorithm and CKFD algorithm on O-ring on-line detection.