面向復雜特征點云的高精高效配準方法

摘" 要： 為滿足工業生產對航空葉片精鑄零件點云配準精度越來越高的需求，并提高配準的速度和精度，提出一種配準方法——組合裁剪迭代最近點（C?TrICP）。首先，使用曲率下采樣算法對點云進行下采樣，較好地保留原始點云的特征；然后，對裁剪迭代最近點（TrICP）算法進行改進，并與迭代最近點（ICP）及TrICP算法進行對比實驗。結果表明：改進算法在各點云模型的配準上均能取得較好的結果，且相比于ICP算法，該算法對Lucy、Bunny及Blade點云的配準效率分別提高了43.03%、43.86%及30.09%，配準精度分別提高了57.90%、99.96%及62.50%，解決了ICP及TrICP算法較為耗時、迭代收斂速度慢的問題，提高了配準精度。最后，將改進算法應用于航空葉片輪廓曲面自測點云的配準上，也取得了較好的配準精度，說明該方法對復雜曲面精密加工的發展具有良好的理論意義和重要的工程應用價值。

關鍵詞： 點云配準； 下采樣； 裁剪迭代最近點算法； 迭代最近點； 收斂速度； 配準精度

中圖分類號： TN249?34" " " " " " " " " " " " " " "文獻標識碼： A" " " " " " " " " " " 文章編號： 1004?373X（2025）04?0181?06

Method of high?precision and high?efficiency registration for complex feature point clouds

CHEN Xin1， ZHANG Dabin1， ZHANG Junfei1， 2， JI Zhengkang3

（1. School of Mechanical Engineering， Guizhou University， Guiyang 550000， China; 2. Guizhou People’s Armed College， Guiyang 550000， China;

3. Guiyang Hangfa Precision Casting Co.， Ltd.， Guiyang 550000， China）

Abstract： In order to meet the increasing demand for point cloud registration accuracy of precision cast parts for aviation blades in industrial production， and to improve the speed and accuracy of registration， a registration method： combined trimmed iterative closest point （C?TrICP）， is proposed. The curvature down?sampling algorithm is used to down?sample the point cloud， which well preserves the characteristics of the original point cloud. The TrICP algorithm is improved， and the comparative experiments with ICP and TrICP algorithms are conducted. The results show that the improved algorithm can realize better results in the registration of each point cloud model， and compared with the ICP algorithm， the registration efficiency of this algorithm in Lucy， Bunny and Blade point cloud is improved by 43.03%， 43.86% and 30.09% respectively， and the registration accuracy is improved by 57.90%， 99.96% and 62.50% respectively. It solves the problem that the ICP and TrICP algorithms are time?consuming and slow in iteration convergence， and can improve the registration accuracy. The improved algorithm is applied to the registration of self?measured point cloud of aviation blade profile surface， and good registration accuracy is also obtained. It has good theoretical significance and important engineering application value for the development of precision machining of complex surfaces.

Keywords： point cloud registration; down?sampling; trimmed iterative closest point algorithm; iterative closest point; convergence rate; registration accuracy

0" 引" 言

近年來，隨著三維掃描技術的日益成熟，三維點云數據的采集變得更加容易，其中還包含了豐富的空間信息。點云配準是點云數據處理的一項基本任務，實現高精度配準對其實際應用具有重要意義[1]。

點云配準的目的是獲得兩個不同點云之間的空間變換[2]，這在三維重建和目標檢測中是至關重要的。三維點云配準技術已廣泛應用于機器視覺[3?4]、三維重建[5?6]、模式識別[7]及柔性制造[8]等領域，也逐漸應用于航空葉片氣膜孔電加工過程位姿偏差的修正?！?br>