Presentation Information
[20a-A308-6][JSAP-Optica Joint Symposia 2023 Invited Talk] Omnidirectional 3D shape measurement using Gabor digital holography and deep neural network
〇Nobukazu Yoshikawa1 (1.Tottori Univ.)
Keywords:
digital holography,omnidirectional 3D measurement,neural network
We have developed an omnidirectional 3D measurement method using Gabor digital holography (DH) and deep neural network (DNN). In this method, con-tour images of an object are obtained from reconstructed images of Gabor holograms captured from all directions. The 3D shape in all directions is restored by combining the contour information from all directions. Since DH is a lens-less imaging technique, accurate 3D shape can be obtained without distortion caused by lens aberrations. We developed a contour re-construction method based on deep learning. A DNN was trained using Gabor holograms created from objects with various shapes and contour information obtained by the edge-enhanced reconstruction method. We demonstrated that the 3D shape can be accurately restored using the contour image obtained by DNN with high accuracy.