To overcome defects of traditional artificial rivets detection

such as low work efficiency and unmanageable precision

a non-contact automatic detection algorithm for multi-rivets based on machine vision is proposed.The improved OTSU algorithm is adopted to segment collected rivet image which effectively reduce erroneous segmentation of stain areas.In order to resolve the impact caused by the randomness of rivets orientation on test process

the minimum enclosing rectangle method is adopted to position the outline spindle of rivet

which conquers the uncertain factors in complex industrial environments.Identify the outline feature points by calculating curvature of each rivet contour point in the both sides of the support region

which effectively reduces weight of noise in curvature calculation and improve the recognition precision.The experiments show that the system has advantageous of high detection precision and low false detection rate

and can meet on-line detection requirements of rivets.