SHINE-Mapping: Large-Scale 3D Mapping Using Sparse Hierarchical Implicit Neural Representations

Accurate mapping of large-scale environments is an essential building block of most outdoor autonomous systems. Challenges of traditional mapping methods include the balance between memory consumption and mapping accuracy. This paper addresses the problem of achieving large-scale 3D reconstruction using implicit representations built from 3D LiDAR measurements. We learn and store implicit features through an octree-based, hierarchical structure, which is sparse and extensible. The implicit features can be turned into signed distance values through a shallow neural network. We leverage binary cross entropy loss to optimize the local features with the 3D measurements as supervision. Based on our implicit representation, we design an incremental mapping system with regularization to tackle the issue of forgetting in continual learning. Our experiments show that our 3D reconstructions are more accurate, complete, and memory-efficient than current state-of-the-art 3D mapping methods.

Automated summary

This paper focuses on achieving large-scale 3D reconstruction from 3D LiDAR measurements using implicit representations. By learning and storing implicit features in a hierarchical structure and converting them into signed distance values through a shallow neural network, the authors propose an incremental mapping system to address the issue of forgetting in continual learning. Experimental results demonstrate that their approach outperforms current state-of-the-art 3D mapping methods in terms of accuracy, completeness, and memory efficiency.