基于双目视觉的涡轮叶片型面特征提取及匹配

Feature Extraction and Matching for Turbine Blade Surface Based on Binocular Vision

提出一种基于双目视觉的测量方法,通过投影文字图案形成叶片表面检测点,采用具有旋转、尺度不变的SIFT算子对左、右摄像机图像进行特征提取及匹配,并通过窄基线剔除法去除匹配后生成的一系列共轭像素点中的误匹配点,实现对涡轮叶片表面点云数据的高精度、低成本提取。实验结果验证了算法的有效性。

In this paper, a precisely and low cost detection method based on stereo vision is presen- ted by projecting some words and patterns on the surface of the turbine blades. After the feature ex- traction and matching of the stereo vision pictures by using the rotation and scale invariant algorithm, namely, SIFT, an narrow baseline approach is fur- ther proposed to remove the false matching points among the acquired conjugate pixels from the SIFT algorithm. Then, the cloud data of turbine blades could be extracted accurately. The experimental re- suits demonstrate the effectiveness of the proposed method.