基于经验小波变换和Beamlet变换的裂纹检测方法  

Crack Detection Based on Empirical Wavelet Transform and Beamlet Transform

作　　者：林哲[1] 蔡恬 王燕锋 Lin Zhe;Cai Tian;Wang Yanfeng(Department of Computer,Shantou Polytechnic,Shantou Guangdong 515078,China;Network and Information Center, Shantou Polytechnic,Shantou Guangdong 515078,China;Architectural Engineering Institute, Zhongyuan University of Technology,Zhengzhou 450007,China)

机构地区：[1]汕头职业技术学院计算机系,广东汕头515078 [2]汕头职业技术学院网络与信息中心,广东汕头515078 [3]中原工学院建筑工程学院,郑州450007

出　　处：《北京联合大学学报》2019年第2期24-31,共8页Journal of Beijing Union University

基　　金：广东省高等学校结构与风洞重点实验室开放课题基金(201601;201803)

摘　　要：在材料完整性检测中,基于计算机视觉的光学检测技术在纹理复杂、噪声较大的背景中准确地检测外部裂纹仍然十分困难。针对这一问题,提出了一种检测裂纹的算法。首先,将裂纹图像自适应地分解为频率分量以得到更好的稀疏表示,然后利用多尺度几何分析技术从重构的近似图像中稳健地提取裂纹的线特征。实验表明,该方法比其他方法能够更准确地检测出图像中的裂纹,在抑制背景纹理和保留裂纹信息这两个性能上取得了较好的综合效果。In integrity inspection of materials,the optical detection technique based on computer still has difficulty to detect external cracks efficiently and accurately on the background with complicated texture and noise. To address this problem,an effective method for robustly detecting crack is presented in this paper. It is characteristic that a crack image is decomposed adaptively to frequency components for better sparse representation,and then linear features of cracks are extracted robustly by multi-scale geometric analysis from the approximately reconstructed image. Experiments on various images show that cracks are detected more accurately by using this method than other methods. It achieves a preferable performance on balancing the competing requirements of suppressing background texture and retaining crack information.

关 键 词：裂纹检测 经验小波变换 BEAMLET变换 稀疏表示 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]