基于显著性检测的风机叶片裂纹提取研究

Research on Extraction of Fan Blade Crack Based on Salient Region Detection

摘要: 在基于数字图像处理技术的风力发电机桨叶故障巡检过程中，为了提取桨叶上的裂纹，结合裂纹与桨叶对比度的特点，文章提出了一种基于显著性区域检测与连通域原理相结合的裂纹提取方法。利用灰度化和基于引导滤波器的图像增强算法对图像进行预处理。然后转变处理前景目标区域的方法，转而分析背景区域的特点，达到把背景区域去除的效果，并与连通域检测标记相结合，最终把裂纹区域提取出来。实验结果表明，基于显著性区域检测的方法能够把风机桨叶上的裂纹较为完整地标记出来。

Abstract: A method based on the combination of the salient region detection and the connected domain detection is proposed in this paper. This method is applied to extract the crack on the blade in the process of the wind turbine blade failure inspection based on digital image processing technology. First, the gray processing and the image enhancement based on the guide filter were used for preprocessing. And then the commonly used processing methods concentrate in the target area was changed into analyzing the characteristics of the background region to remove the background area. Finally, the salient region detection and the connected domain detection were combined to extract the crack area. The test results show that this method can label the cracks from the blade comparatively completely.