基于SIFT算法及阈值筛选的点云配准技术研究  

Research on Point Cloud Registration Technology Based on SIFT Algorithm and Threshold Filter

作　　者：顾旭波[1] 张永举[1] 张健 吴良成 郭玲[2] 

机构地区：[1]江苏省特种设备安全监督检验研究院吴江分院,江苏苏州215200 [2]南京理工大学自动化学院,南京210094

出　　处：《计算机测量与控制》2017年第12期247-250,共4页Computer Measurement &Control

基　　金：江苏省质监局2016年度科研项目(KJ168357)

摘　　要：随着三维测量技术应用领域的逐渐拓宽,点云数据处理技术的需求日益迫切,而多视点点云配准,是其中的基础技术环节;在此针对传统ICP算法鲁棒性差、对迭代初值敏感、计算效率低等缺点,提出一种SIFT算法与阈值筛选相结合的点云配准算法;在参考点云和待配准点云中,通过计算SIFT关键点及各点主曲率,获得初始匹配点集;然后根据相似三角形阈值和法向量夹角阈值,进一步优化点对间的旋转平移关系;实验结果证明,相对于传统算法,改进算法能够以更短的时间来获得准确的配准效果,并且其自动化程度高以及能有效提高点云配准的效率和精度。With widening application of 3D measurement technology, the demand for cloud point data processing technology is becoming more and more urgent, and the multi view point cloud registration is one of the fundamental technologies. A point cloud registration algo- rithm combining SIFT algorithm with threshold selection is proposed to overcome the disadvantages of the traditional ICP algorithm, such as poor robustness, sensitive to iterative initial value and low computational efficiency. Firstly, the initial matching set between the reference point cloud and the point cloud to be registered is obtained by calculating the SIFT key points and their main curvatures~ then rotation and translation between corresponding are optimized based on the similar triangle threshold and the vector angle threshold. Experiments show that, compared with the traditional algorithm, the improved algorithm can achieve accurate registration results in shorter time, and it is high- ly automated and can effectively improve the efficiency and accuracy of point cloud registration.

关 键 词：关键点 SIFT算子 点云配准 相似三角形 

分 类 号：TP391.7[自动化与计算机技术—计算机应用技术]