Computer Science

CLC number: TP391.41

Adaptive Fast Point Feature Histogram Registration with Dynamic Downsampling

融合动态降采样的自适应FPFH配准

¹ School of Electronic and Electrical Engineering, Shanghai University of Engineering Science, Shanghai 201620, China (上海工程技术大学　电子电气工程学院，上海　201620)
² School of Optical-Electrical and Computer Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China (上海理工大学　光电信息与计算机工程学院，上海　200093)
³ School of Chemistry and Chemical Engineering, Shanghai University of Engineering Science, Shanghai 201620, China (上海工程技术大学　化学与化工学院，上海　201620)

^† Corresponding author. E-mail: This email address is being protected from spambots. You need JavaScript enabled to view it.

Received: 28 July 2025

Abstract

In the context of rapid advancements in laser scanning technology, the scale of point cloud data acquisition has increased significantly. Existing registration algorithms face escalating computational costs when processing medium- and large-scale point clouds, which severely impairs registration efficiency. To address this issue, we propose an enhanced registration framework based on Fast Point Feature Histogram (FPFH) that incorporates dynamic downsampling and adaptive neighborhood optimization. Specifically, the global average point spacing is first computed to guide the downsampling process. Subsequently, the FPFH neighborhood radius is dynamically adjusted according to this global average spacing. Experimental results on two datasets indicate that our approach maintains robust accuracy while cutting computation time by approximately 55% compared with the standard FPFH algorithm for medium- and large-scale clouds. Furthermore, compared with the traditional 3D Shape Context (3DSC) algorithm, our approach requires less than 10% of the processing time and achieves significantly higher registration precision.

摘要

随着激光扫描技术的迅猛发展，点云数据的采集规模呈显著增长态势。现有的配准算法在处理中大规模点云时，面临计算开销急剧攀升的挑战，严重制约了配准效率。针对这一问题，本文提出了一种基于快速点特征直方图(FPFH)的改进型配准框架，该框架融合了动态降采样与自适应邻域优化策略。具体而言，算法首先计算全局平均点间距，以此为基准指导降采样过程；随后，依据该全局平均间距动态调整FPFH的邻域半径。在两个数据集上的实验结果表明，相较于标准FPFH算法，该方法在处理中大规模点云时，在保持稳健精度的前提下，将计算时间缩减了约55%。此外，与传统的3D形状上下文(3DSC)算法相比，本文方法仅需不到10%的计算时间，并显著提升了配准精度。