Abstract We describe an algorithm for generating shape primitives from images of machined parts. The algorithm uses the properties of the Hough transform and its inverse. The images are captured as 256 × 256 pixel arrays, with 256 gray levels. Edge detection is performed using a Sobel transform and the binarized, edge-detected image is Hough transformed. The Hough plane is searched for the peaks due to straight lines in the image, and the bands due to curved lines. On output, straight lines in the image are represented in parametric form, and each curved edge is represented by a set of short straight-line segments. The implementation is robust enough to handle noisy data and will cope with arbitrarily curved edges.
[1]
James F. Boyce,et al.
The Radon transform and its application to shape parametrization in machine vision
,
1987,
Image Vis. Comput..
[2]
Dana H. Ballard,et al.
Generalizing the Hough transform to detect arbitrary shapes
,
1981,
Pattern Recognit..
[3]
David Casasent,et al.
Curved object location by Hough transformations and inversions
,
1987,
Pattern Recognit..
[4]
King-Sun Fu,et al.
Syntactic Pattern Recognition And Applications
,
1968
.
[5]
Mark B. Sandler,et al.
An Efficient Radon Transform
,
1988,
Pattern Recognition.