Objective Light field imaging technology is an important research direction in computational imaging, combining optical design, optical principles, and computational processing to break through the technical barriers of traditional optical design to a certain extent. Compared with traditional cameras, light field cameras introduce a micro-lens array (MLA) between the main objective lens and the detector, and can obtain both positional and angular information with just one shot. Currently, light field camera 3D imaging technology is mainly applied to close-range scenes, such as biomedicine, high-end industrial inspection, and particle tracking, while there are very few applications in long-range 3D imaging. Meanwhile, 3D imaging technology based on light field camera is an integrated computational imaging technology of front-end optical systems and back-end information processing. At present, there are many studies on the back-end processing algorithms of light field camera, but few reports on the impact of front-end optical system parameters. Therefore, this paper investigates the influence of front-end optical system parameters of light field cameras on longitudinal resolution in long-range 3D imaging and conducts relevant experiments for verification.

Methods Based on the principle of multi-view vision and the imaging principle of focused light field cameras, this paper studies the virtual depth (Fig.1) and the matching relationship between the F/# of the main objective lens and that of the microlenses (Fig.2), establishes a longitudinal resolution model for light field cameras (Fig.3), and derives the mathematical expression of longitudinal resolution based on optical parameters.

Results and Discussions This paper analyzes the impact of optical parameters in the front-end optical system, such as virtual depth \nu , lens distance coefficient k maximum number of visible microlenses N, focal length of the main objective lens, F/# of the main objective lens, and F/# of the microlenses, on the longitudinal resolution of light field cameras (Fig.5-Fig.10). Based on the above analysis, a light field camera with a main objective lens focal length of 900 mm (Fig.11), F/#=6.4, and a full-field MTF better than 0.5 at a certain spatial frequency is designed (Fig.12). The microlenses are arranged in a hexagonal pattern with a focal length of 1.4 mm and are placed 0.7 mm away from the detector. Relevant experiments are also conducted (Fig.13).

Conclusions Experiments on long-range 3D imaging were conducted for six different targets ranging from 640 m to 1510 m. The experimental results show that the longitudinal resolution accuracy was 4.8 m at 0.64 km and 10.39 m at 1.07 km (Fig.14-Fig.19). The light field distance calculation was basically consistent with the measured data (Fig.20). Meanwhile, the experimental calculation accuracy was in line with the theoretical analysis accuracy (Fig.21).