基于曲波域移动平行窗的受电弓滑板裂纹识别  

Pantograph Slipper Cracks Identification Based on Translational Parallel Window in Curvelet Transform Domain

作　　者：陈坤峰[1] 刘志刚[1] 韩志伟[1] 何士玉[1] 

机构地区：[1]西南交通大学电气工程学院,四川成都610031

出　　处：《铁道学报》2012年第10期43-47,共5页Journal of the China Railway Society

基　　金：国家自然科学基金(U1134205;51007074);教育部新世纪优秀人才支持计划(NET-08-0825);铁道部科技研究开发计划(2011J016-B);中央高校科研业务费专项资金(SWJTUCX141)

摘　　要：针对受电弓滑板非接触式裂纹故障检测问题,提出一种基于曲波域移动平行窗的受电弓滑板裂纹识别算法。算法利用曲波变换多方向性及各向异性特点,对受电弓滑板图像点状、线性和非线性特征进行分类。在曲波分解方向矩阵中使用移动平行窗口并计算窗口能量值,依据能量值区分线性平行接缝与背景噪声、螺钉和刮痕等其他非线性图像特征,最终获取滑板裂纹信息。实验结果表明,本文算法能有效地分类滑板图像特征,较准确地检测并定位滑板裂纹故障,识别率达到94.1%。Aiming at solving the problem of pantograph slipper cracks non-contact detection,a novel algorithm based on the translational parallel window in the curvelet transform domain was proposed.Utilizing the multi-direction and anisotropy properties of curvelet transform,the spot,linear and nonlinear characteristics of the pantograph slipper images were classified.Translational parallel windows were used in the direction matrixes of curvelet decomposition,and then the energy values of translational windows were calculated.The linear parallel joints and other nonlinear images such as background noises,rivets,scratches,and cracks were distinguished from one another according to the energy values of the curvelet coefficients,so the slipper cracks information would be obtained finally.The experimental results show that the proposed algorithm can classify the characteristics of the slide images effectively,detect and identify the slipper cracks accurately,and the identification rate can reach 94.1%.

关 键 词：受电弓滑板 曲波变换 裂纹提取 移动平行窗 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]