Magnetic　rings　play　an　important　role　in　the　manufacture　of　high-tech　products　in　the　fields　of　aerospace,　military　electronics,　and　automobiles,　whose　manual　detection　of　surface　defects　is　costly,　inefficient,　and　low-precision.　In　order　to　solve　these　problems,　this　paper　proposes　a　magnetic　ring　multi-defective　stereo　detection　system　based　on　group　vision　technology,　which　completes　the　detection　of　each　surface　of　three-dimensional　magnetic　ring.　The　detection　algorithm　has　a　recognition　rate　of　100%　for　defects　such　as　cracks,　adhesions,　hanger　adhesions,　pitting,　and　a　recognition　rate　of　more　than　90%　for　defects　such　as　fibers　and　foreign　objects.　The　average　working　speed　of　the　system　is　over　120　per　minutes　and　the　defect　detection　accuracy　is　within　0.05　mm.　The　accuracy　and　efficiency　of　the　system　can　completely　meet　the　requirements　of　actual　production.　©　2018　IEEE.