利用改进的SIFT算法检测桥梁拉索表面缺陷

设计了一种分布式机器视觉检测系统对斜拉桥拉索表面的损伤缺陷进行自动无损检测。该系统采用4个CCD摄像头分布在拉索表面一周获取图像,一个缺陷有可能分布在几幅图像中。为了快速有效地获得完整的拉索表面缺陷,本文提出了改进的尺度不变特征变换(SIFT)特征匹配算法对缺陷图像进行自动拼接。首先,采用简洁有效的Harris算子提取特征点;然后,根据检测系统采集的缺陷图像的特点,简化SIFT算子的特征点主方向分配和匹配图像旋转等算法步骤,对特征点进行描述和匹配;最后,融合匹配图像,得到相对完整的缺陷图像。实验结果表明,利用改进的SIFT算法对拉索表面缺陷图像进行自动拼接,降低了算法的复杂度,提高了桥梁拉索表面缺陷检测的完整性。

Using Improved SIFT Algorithm to Implement Surface Defects Detection for Bridge Cable

In order to realize automatic nondestructive testing for surface cable damage on a cable-stayed bridge，a distributed machine vision system was developed. It uses four cameras to acquire images around the cable surface. Surface defection may be distributed in several images. An improved scale invariant feature transform(SIFT) feature matching algorithm for image mosaicing is proposed to real time processing to obtain a whole defect effectively. First，feature points are extracted by a Harris operator. Second，according to defect images collected by the system，the steps of the SIFT operator such as the distribution of the main direction for the matching feature points and the matcking image rotation is simplified. The simplified SIFT operator is employed to describe the feature points and match the images. Finally，image fusion is implemented and a complete image of a defect is obtained. Experimental results show that the algorithm complexity is greatly reduced and improves detection integrity for surface cable defects using our improved SIFT to automatically stitch the defect images together.