Three-Dimensional Reconstruction Method for Bionic Compound-Eye System Based on MVSNet Network

In practical scenarios, when shooting conditions are limited, high efficiency of image and success rate 3D reconstruction required. To achieve the application bionic compound eyes in small portable devices for reconstruction, auto-navigation, obstacle avoidance, a deep learning method using compound-eye system with partial-overlap fields was studied. We used to capture images target scene, then restored camera parameter matrix by solving PnP problem. Considering unique characteristics system, we designed neural network based on MVSNet structure, named CES-MVSNet. fed captured parameters trained network, which can generate results good integrity precision. traditional multi-view geometric networks difference between effects two methods analyzed. The reliability proved.