基于多分辨率的动车车轴图像快速配准算法  

Multi resolution based fast registration algorithm for emu axle images

作　　者：刘志[1] 张若宸 

机构地区：[1]北方工业大学电子信息工程学院,北京100144

出　　处：《计算机工程与设计》2016年第10期2754-2758,共5页Computer Engineering and Design

摘　　要：为降低高分辨率图像配准的高耗时,优化动车组运行故障动态图像检测系统(trouble of moving emu detection system,TEDS),设计一种基于多分辨率图像的动车车轴图像快速配准算法。利用SURF(speeded up robust features)算子的旋转缩放不变性在低分辨率图像中进行粗配准,对经典的HOG(histogram of oriented gradient)算法进行优化,在高分辨率图像中利用优化了的HOG算法进行筛选及精确配准。实验结果表明,该算法提高了配准性能,加快了配准的速度。To reduce the high cost of time of high-resolution image registration,the TEDS was optimized,and a multi resolution based fast image registration algorithm was proposed.The rotation and scale invariant of SURF was utilized for coarse registration in low resolution layer.The classical HOG was improved,and the improved HOG was used for exact registration in high resolution layer.Experimental results show that the proposed algorithm can improve the registration accuracy and reduce registration time.

关 键 词：图像配准 多分辨率 HOG SURF TEDS 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]