In applications such as scene reconstruction and odometry, accurate matching associations for 3D lines are crucial. Real-world scenes introduce inconsistencies due to variations in perspective, leading to non-overlapping data acting as noise. Accurately matching partially overlapping sets of 3D lines becomes challenging, potentially resulting in failed scene reconstruction and erroneous positioning. Prior approaches relied on traditional Iterative Closest Line (ICL) methods, involving iterative calculations and sensitivity to initial poses, and were prone to matching failures in low-overlap rate data and singular pattern scenes. Existing 3D line matching networks either did not consider the noise in 3D line collections or failed to retain more valid matching pairs, while these models often require a larger number of parameters and inference time. To address these issues, this paper proposes MEG-Net, a Plücker line matching network guided by the existence of matches. It leverages the rich geometric characteristics of 3D lines represented as Plücker lines, enhancing feature robustness. By guiding the model to handle noisy data through match existence guidance, it improves the model’s performance on partially overlapping 3D line data. Experiments on indoor and outdoor datasets and out-of-distribution (OOD) datasets demonstrate that MEG-Net outperforms traditional methods and baseline models in 3D line matching, with better scalability and noise robustness, achieving state-of-the-art results. Additionally, we propose an innovative knowledge distillation method based on matching matrices, training a more efficient MEG-Net mini student model with approximately 70% fewer parameters and MACs (Multiply Accumulate Operations), while maintaining superior performance and faster inference speeds on indoor datasets.

中文翻译：

---

基于匹配存在性引导和知识蒸馏的3D线匹配网络


在场景重建和里程计等应用中，3D 线条的精确匹配关联至关重要。现实世界的场景由于视角的变化而引入不一致，导致不重叠的数据充当噪声。准确匹配部分重叠的 3D 线集变得具有挑战性，可能导致场景重建失败和定位错误。先前的方法依赖于传统的迭代最近线（ICL）方法，涉及迭代计算和对初始姿态的敏感性，并且在低重叠率数据和奇异模式场景中容易出现匹配失败。现有的3D线匹配网络要么没有考虑3D线集合中的噪声，要么未能保留更多有效的匹配对，而这些模型通常需要更多的参数和推理时间。为了解决这些问题，本文提出了 MEG-Net，一种由匹配的存在性引导的 Plücker 线匹配网络。它利用以 Plücker 线表示的 3D 线的丰富几何特征，增强了特征的稳健性。通过匹配存在指导来引导模型处理噪声数据，提高了模型在部分重叠的 3D 线数据上的性能。对室内和室外数据集以及分布外（OOD）数据集的实验表明，MEG-Net 在 3D 线匹配方面优于传统方法和基线模型，具有更好的可扩展性和噪声鲁棒性，取得了最先进的结果。 此外，我们提出了一种基于匹配矩阵的创新知识蒸馏方法，以减少约 70% 的参数和 MAC（乘法累加运算）来训练更高效的 MEG-Net 迷你学生模型，同时在室内数据集上保持卓越的性能和更快的推理速度。