新型三维形貌测量系统SIFT算法改进研究  

Improved SIFT Algorithm Based on a New Measurement System of Large Scale 3D Shape

作　　者：刘常杰[1] 黄东晓[1] 杨凌辉[1] 

机构地区：[1]天津大学精密测试技术及仪器国家重点实验室,天津300072

出　　处：《传感技术学报》2014年第11期1500-1505,共6页Chinese Journal of Sensors and Actuators

基　　金：国家863计划项目(2012AA041205)

摘　　要：将测绘领域的三线阵摄影测量原理应用于运动物体三维形貌测量,提出一种基于三线阵CCD的新型三维形貌测量方法。针对新型测量系统获取的三线阵CCD影像同名点匹配问题,研究了常用的影像匹配算法,并选用SIFT匹配算法。由于传统SIFT匹配算法处理大尺寸图像耗时长、提取匹配点数少等不足,对算法进行优化,并提出基于新型大尺寸三维形貌测量系统的匹配搜索策略,确定最优匹配阈值,最后通过对比实验进行验证。实验表明,改进的算法能够解决匹配影像视角变化等问题,缩短算法处理时间,增加匹配点数,提高算法性能。The tri-linear CCDs photogrammetry using in surveying was applied to large scale 3-D shape measurement. To pick out the homologous image points from tri-linear CCDs images,SIFT algorithm was used final-ly. In order to against deficiencies of traditional SIFT,such as time consuming,a new improved SIFT algorithm was proposed. A searching strategy and threshold based on new system were given to meet“real-time”,verified by the experiments. The experiments show that the improved SIFT can cut down the time-consuming and increase the num-ber of matching point-pair and performance of algorithm.

关 键 词：三维形貌测量 三线阵CCD SIFT算法 影像匹配 

分 类 号：TP391[自动化与计算机技术—计算机应用技术] TH741[自动化与计算机技术—计算机科学与技术]