A clustering based hierarchical image feature location method for complex shaped work-piece

This paper addresses an image feature location method for complex shaped work-piece. The “rough estimation first and then precise location” hierarchical feature location method is proposed to roughly estimate the work-piece´s pose first and then precisely locate it in 2D image space. The proposed method has the following main steps: (1) simple features are detected first then more complex features are examined later, using the locations of the previously found features; (2) an improved local adaptive threshold algorithm is used to binarize the work-piece´s gray image; (3) a size adjustable binary template is designed to match hole features; (4) a work-piece is finally located in image space according to its CAD model using the nearest neighbor clustering algorithm. The method has been tested on a sort of complex shaped automobile engine cylinder head as an example and the location results are satisfactory.