• Journal of the European Optical Society-Rapid Publications
  • Vol. 20, Issue 1, 2024032 (2024)
Stéphane Cuenat, Jesús E. Brito Carcaño, Belal Ahmad, Patrick Sandoz..., Raphaël Couturier, Guillaume J. Laurent and Maxime Jacquot*|Show fewer author(s)
DOI: 10.1051/jeos/2024032 Cite this Article
Stéphane Cuenat, Jesús E. Brito Carcaño, Belal Ahmad, Patrick Sandoz, Raphaël Couturier, Guillaume J. Laurent, Maxime Jacquot. Digital holographic microscopy applied to 3D computer micro-vision by using deep neural networks[J]. Journal of the European Optical Society-Rapid Publications, 2024, 20(1): 2024032 Copy Citation Text show less
(a) Lyncee-tec DHM observing a micro-structured pattern moved by a hexapod stage. (b) A typical experimental hologram of a pseudo-periodic pattern that allow 3D pose measurement [2]. Image reconstruction (c) in amplitude and (d) in phase at a numerical in-focus distance of 185 μm.
Fig. 1. (a) Lyncee-tec DHM observing a micro-structured pattern moved by a hexapod stage. (b) A typical experimental hologram of a pseudo-periodic pattern that allow 3D pose measurement [2]. Image reconstruction (c) in amplitude and (d) in phase at a numerical in-focus distance of 185 μm.
(a–c) Thumbnail reconstruction. (d–f) Assess the distance Z. (a) A ROI of 768 × 768 is cropped from the hologram at a fixed position. (b) XY Model (based on a UNet like model). (c) The reconstructed thumbnail of 64 × 64 pixels. (d) A ROI of 128 × 128 is randomly cropped from the hologram space. (e) Z model based on an adapted version of a GedankenNet model [3]. (f) The distance Z.
Fig. 2. (a–c) Thumbnail reconstruction. (d–f) Assess the distance Z. (a) A ROI of 768 × 768 is cropped from the hologram at a fixed position. (b) XY Model (based on a UNet like model). (c) The reconstructed thumbnail of 64 × 64 pixels. (d) A ROI of 128 × 128 is randomly cropped from the hologram space. (e) Z model based on an adapted version of a GedankenNet model [3]. (f) The distance Z.
(a) Outliers (in red), simulated (in blue) and estimated (in green) trajectory in the 3D space. (b) Z and X-Y errors in μm (absolute difference and L2-norm). The Z error is mostly below an error of 1 μm (red dashed line).
Fig. 3. (a) Outliers (in red), simulated (in blue) and estimated (in green) trajectory in the 3D space. (b) Z and X-Y errors in μm (absolute difference and L2-norm). The Z error is mostly below an error of 1 μm (red dashed line).
Matching rate associated to each 3D pose (red: outliers, green: right 3D poses).
Fig. 4. Matching rate associated to each 3D pose (red: outliers, green: right 3D poses).
Stéphane Cuenat, Jesús E. Brito Carcaño, Belal Ahmad, Patrick Sandoz, Raphaël Couturier, Guillaume J. Laurent, Maxime Jacquot. Digital holographic microscopy applied to 3D computer micro-vision by using deep neural networks[J]. Journal of the European Optical Society-Rapid Publications, 2024, 20(1): 2024032
Download Citation