Three-dimensional measurement of precise shaft parts based on line structured light and deep learning

Line structured light is widely used in three-dimensional measurement of metal parts. The large curvature and specular reflection on the surface of precise shaft parts make the intensity of laser stripes captured at some different positions too weak or too strong to extract the center accurately. To extract the center of the stripes from the disturbed image, a center extraction algorithm based on deep learning is proposed. The deep neural network with good feature learning ability can extract the overall distribution and bending characteristics of laser stripes. The region of laser stripe is accurately detected and segmented from the interference. Then the center of the segmented stripe can be well extracted. Experiment verifies the noise suppression ability of the proposed method. Three-dimensional measurement of the precise shaft parts is implemented with light plane calibration. The diameter of the shaft parts is calculated and the measurement error is less than 0.017 mm.