3D localization of point source is widely used in many fields, such as bioimaging and autonomous driving fields. However, the localization is hard to perform under scattering conditions because of the diffuse effect of the scattering. We propose a novel method for 3D localization of point source under scattering conditions based on light field imaging and deep learning by only one shot. First, we introduce the description of the point source in a light field wise and how to localize a point source by its light field. On the basis, we elaborate on the effect of scattering on a light field and how to retrieve the location of a point source from a light field with scattering. Then, the effect of aberration on a light field will be introduced. We also build an artificial scene and a deep learning framework to perform a 3D localization practically, and the feasibility and accuracy of our method have been evaluated.
Deformable mirror (DM) is a flexible wavefront modulator with a changeable surface. It is traditionally adopted in adaptive optical system for aberration correction. Recently applications in zoom imaging system and interferometer for freeform measurement have been proposed because the improvement in fabrication technique makes larger stroke amount and faster response possible. The order and accuracy of aberration correction are typical wavefront correction characteristics of DMs. Due to the non-linearity, hysteresis and creep characteristic of piezoelectric ceramics, accurate control of piezoelectric type DM remains a challenge. Generally, the surface shape of a DM is changed by altering the voltages applied to different actuators below the DM film. And the shape of the DM can be fitted with Zernike polynomial to better characterize the aberration. So accurate control of the DM surface shape requires a relationship between the control voltage vector and the Zernike coefficients of the surface shape. We adopt neural network for the foundation of the relationship. 3000 set of control-voltage-vector and Zernike-coefficient pairs are experimentally collected based on the data measured with an interferometer and fitted with Zernike polynomials. The neural network is constructed and trained, and the control voltage vectors of new surface shapes can be retrieved with the network. The accuracy of shape realization is finally demonstrated by comparison between measured and predicted voltages.
Micro-structured array is a crucial optical element with wide range of applications. The optical performance of microstructured array is determined by feature sizes of array, such as diameter, depth and the uniformity across the whole array. Those sizes can be directed retrieved from the 3D profile. We propose a 3D profile measurement system based on light field microscope, which is promising in achieving fast data acquisition by one shot. We propose the principle of measurement, develop the algorithm for focus stack calculation and 3D reconstruction. Preliminary experiments suggest the prospects and challenges.