基于非下采样Shearlet变换的磁瓦表面裂纹检测

发布时间：2018-03-19 03:14

本文选题：磁瓦　切入点：非下采样Shearlet变换　出处：《农业机械学报》2017年03期 　论文类型：期刊论文

【摘要】：针对磁瓦表面裂纹缺陷图像背景不均匀、对比度低和存在纹理干扰等特点,提出了一种基于非下采样Shearlet变换(Nonsubsampled Shearlet transform,NSST)的裂纹检测方法。首先对原始图像进行多尺度、多方向NSST分解,得到一个低频子带和多个高频子带,然后利用各向异性扩散和改进的γ增强方法对高频子带进行滤波和增强;同时利用二维高斯函数对低频子带进行卷积操作来构造高斯多尺度空间,估计出图像的主要背景,并通过背景差法得到均匀的低频目标图像。最后通过重构NSST系数得到去噪和增强后的均匀目标图像,利用自适应阈值分割和区域连通法提取裂纹缺陷。实验结果表明,所提方法检测准确率达92.5%,优于基于形态学滤波方法、基于Curvelet变换方法和基于Shearlet变换方法等现有磁瓦表面裂纹检测方法。
[Abstract]:In view of the characteristics of uneven background, low contrast and texture interference in the surface crack defect image of magnetic tile, a new crack detection method based on non-downsampling Shearlet transform nonsubsampled Shearlet transform (NSST) is proposed. A low frequency subband and a plurality of high frequency subbands are obtained by multidirectional NSST decomposition, and then the high frequency subbands are filtered and enhanced by anisotropic diffusion and improved 纬 enhancement methods. At the same time, we use the two-dimensional Gao Si function to convolution the low-frequency subband to construct the Gao Si multi-scale space, and estimate the main background of the image. The uniform low frequency target image is obtained by background difference method. Finally, the uniform target image after denoising and enhancement is obtained by reconstruction of NSST coefficient, and the crack defects are extracted by adaptive threshold segmentation and region connectivity method. The experimental results show that, The accuracy of the proposed method is 92.5, which is superior to the existing methods such as morphological filtering, Curvelet transform and Shearlet transform.
【作者单位】：四川大学制造科学与工程学院;昆明理工大学机电工程学院;
【基金】：“十二五”国家科技支撑计划项目(2015BAF27B01) 四川省科技支撑计划项目(2016GZ0160)
【分类号】：TP391.41;TM351

【相似文献】