Optical lens installation and adjustment pose error identification method based on deep learning

• Main Authors: LI CHAOJIANG, JIN XIN, ZHU RONGQUAN, ZOU DONGYI, LIU DEPIAO, XIE XIANGZHI, ZHANG ZHIJING
• Format: Patent
• Language: Chinese; English
• Published: 09.02.2024
• Subjects: CALCULATING; COMPUTING; COUNTING; ELECTRIC DIGITAL DATA PROCESSING; MEASURING; MEASURING ANGLES; MEASURING AREAS; MEASURING IRREGULARITIES OF SURFACES OR CONTOURS; MEASURING LENGTH, THICKNESS OR SIMILAR LINEARDIMENSIONS; OPTICAL ELEMENTS, SYSTEMS, OR APPARATUS; OPTICS; PHYSICS; TESTING

The invention discloses an optical lens installation and adjustment pose error identification method based on deep learning, and belongs to the technical field of reflective optical lens assembly. The pose error identification method comprises the following steps: 1, constructing a lens adjustment light path model, and placing a lens to be adjusted at a theoretical design position; 2, setting a lens installation and adjustment pose error and carrying out wave aberration Zernike analysis to obtain a large number of installation and adjustment pose error-Zernike coefficient data sets; 3, based on a deep learning method, constructing a nonlinear prediction model of the Zernike coefficient and the installation and adjustment pose error; and 4, detecting a Zernike coefficient by using the laser interferometer during actual adjustment, performing pose error prediction by calling the deep learning model, and controlling the pose of the lens through the motion execution assembly to complete the adjustment. According