ERASOR++ introduces Height Coding Descriptor, Height Stack Test, Ground Layer Test, and Surrounding Points Test to enhance dynamic object removal. The system outperforms previous methods in accuracy and speed, addressing challenges in map building with dynamic objects present.
The paper discusses the importance of accurate mapping in automatic systems and the impact of dynamic objects on map distortion. Various methods are compared for dynamic object removal from LiDAR point clouds. The proposed ERASOR++ method demonstrates superior performance through novel techniques like HCD and HST.
Key contributions include a novel representation of descriptors, comprehensive test methods for evaluating dynamic bins, and improved performance over previous work. Ablation experiments confirm the effectiveness of each part in reducing bad removals and enhancing system generalization.
Experimental results show significant improvements in Preservation Rate and F1 score while maintaining comparable speed to previous methods. Future work aims to further refine descriptors for broader applications beyond dynamic object removal.
To Another Language
from source content
arxiv.org
Key Insights Distilled From
by Jiabao Zhang... at arxiv.org 03-11-2024
https://arxiv.org/pdf/2403.05019.pdfDeeper Inquiries