Line Extraction of industrial Parts Based on Least Square Template Matching

High precision three-dimensional reconstruction and measurement of industrial objects by digital imagery is one of the most active research fields in Computer Vision and Digital Photogrammetry. Due to most of industrial sheet-metal parts are made up of linear features, it is important to extract lines from 2D images accurately. Meanwhile, quality of lines will further to decides on precision of reconstruction of objects. In this paper, edge detector firstly is performed to extract edges of parts. Edge detection is an important pre-processing step in image analysis. Then, Hough transform is utilized to form an initial line segments. Lastly, least square template matching will further to extract line segments, which was used to get sub-pixel precision. Therefore, principle and methods of line segment least squares template matching to extract lines from imagery are described. Corresponding experiments and results are demonstrated with the conclusion.