Abstract: Bearing surface defects can be divided into two classes, one is the defect of gray-scale lower than the target and another is the defect of gray-scale higer than the target.Aiming at the limitation that single-threshold segmentation algorithm can only detect the first defect and the situation that multi-threshold segmentation algorithm can only detect the second, a new rapid detection method on the bearing surface defects is introduced based on multiple-OSTU.The complexity of the algorithm is smaller than multi-threshold one, and according to the characteristics of the bearing image, it narrows the range of threshold values.So it satisfies the need of on-line detection of bearing surface defects.