基于小波变换和立体视觉的发动机内窥研究

Study of engine endoscopic system based on wavelet transformation and stereo vision

提出一种基于小波变换及立体视觉技术相结合的发动机内窥检测系统。系统由平行光轴双CCD内窥镜及内窥图像处理系统两部分组成,内窥图像处理子系统接收平行光轴双CCD内窥镜采集到的航空发动机内表面图像,对可能存在的损伤缺陷进行识别,并计算出损伤长度、面积、深度等数据。内窥图像处理子系统中,针对发动机内窥图像特点,采用Gauss小波变换方法对图像进行特征提取、立体匹配;利用双目立体视觉技术,三维重建出损伤区域的原貌。本检测系统可以弥补传统检测系统的缺陷,提高检测精度,缩短检测时间,从而保障航空发动机的安全运行。实验结果表明,系统准确有效。

Presents a visual engine fault diagnosing system based on wavelet transformation combined with stereo vision technology. Image processing subsystem receives images from binocular CCD endoscopy and diagnoses faults that occur on the inner surface of aeroengine cavity or turbine vines. There are a few kinds of endoscopic systems applied in practice, but they can not derive adequate data from images direcly such as depth of the pit areas. That is inefficient. Therefore, our diagnosing system ales into account Gauss wavelet transformation to extract features and makes a stereo match of the left and right images. Further more, the author reconstructed the original 3D model of the defect area, which provides more precise intuitionistic information to identify the classification of the defects. This system compensates limitation of traditional diagnosing system, improves diagnosis accuracy, shortens detection time, and assures the safety of engine running.