Towards End-to-End Neuromorphic Voxel-based 3D Object Reconstruction Without Physical Priors

Neuromorphic cameras, also known as event cameras, are asynchronousbrightness-change sensors that can capture extremely fast motion withoutsuffering from motion blur, making them particularly promising for 3Dreconstruction in extreme environments. However, existing research on 3Dreconstruction using monocular neuromorphic cameras is limited, and most of themethods rely on estimating physical priors and employ complex multi-steppipelines. In this work, we propose an end-to-end method for dense voxel 3Dreconstruction using neuromorphic cameras that eliminates the need to estimatephysical priors. Our method incorporates a novel event representation toenhance edge features, enabling the proposed feature-enhancement model to learnmore effectively. Additionally, we introduced Optimal Binarization ThresholdSelection Principle as a guideline for future related work, using the optimalreconstruction results achieved with threshold optimization as the benchmark.Our method achieves a 54.6% improvement in reconstruction accuracy compared tothe baseline method.