The research on line features extraction from images of industrial parts based on CAD data

For features extraction of image measurement and reconstruction from industrial parts images, this paper puts forward a mathematical model guided by CAD designed data which can detect image lines with high accuracy. Firstly, based on CAD data getting line initial values by projecting object line features through the conversion of 3D CAD coordinate and image, and then extracting a sequence of edge points by least squares template matching (LSTM) and acquiring image line equations by the points based on principle of least squares fitting, finally precisely obtaining the contours of industrial parts by solving the intersections of image intersecting lines. By using the algorithm on real images of industrial parts we make a line feature extraction experiment that proved this algorithm was given a high automation extraction, strong stability, and can be used to the image measurement and reconstruction of industrial parts with high accuracy.