融合Hough直线检测和Grab-cut的风机叶片自适应分割方法

完整的边缘信息对风力发电机叶片的边缘缺陷检测至关重要,但由于户外采集的风机叶片图像背景复杂多样,现有图像分割算法的分割精度不足,无法保证边缘缺陷的完整性。因此,提出一种基于图像边缘特征与颜色信息的自适应图像分割方法实现风机叶片边缘检测。首先,使用Hough直线检测初步定位叶片直线边缘;然后,在目标区域应用基于Otus阈值分割和形态学运算的Grab-cut算法,实现叶片图像的自适应分割。采用无人机采集多个场景的图像作为测试样本,对分割方法与其他方法进行定性和定量对比分析。结果表明,该方法能自适应且准确地实现风机叶片图像分割,并保留边缘缺陷的完整性,其边缘覆盖率(0.971 7)和边界位移误差(3.040 3)指标均优于其他方法,对风机叶片的边缘缺陷检测具有重大潜在应用价值。

Adaptive segmentation method for wind turbine blades combining Hough line detection and Grab-cut algorithm

Complete edge information is essential to the detection of edge defects of turbine wind turbine blades. Due to the complex and diverse background of wind turbine blade ( WTB) images, the existing image segmentation algorithms have insufficient segmentation accuracy and cannot guarantee the integrity of edge defects. Therefore, an adaptive image segmentation based on image edge features and color information is proposed for the edge detection of WTBs. Firstly, Hough line detection is used to detect the blade edge at straight lines. Secondly, the Graph-cut algorithm based on Otus threshold segmentation and morphological operations is applied to adaptively separate the blade target areas. Finally, comparative experiments are carried out with a lot of image samples captured under various scenes. The results show that higher edge coverage and lower boundary displacement errors of the WTB image segmentation with 0. 971 7 and 3. 040 3 are obtained by the proposed method as compared with other image segmentation methods. The proposed approach has potential application value for the edge defect detection of WTBs.