摘要
利用小波多尺度特性提取图像边缘是目前研究热点之一。铁谱图像磨粒边缘作为铁谱图像中最基本的特征,为铁谱图像特征提取提供了非常有价值的重要特征参数,在基于铁谱分析的机器故障诊断技术中有着重要的地位。将第二代提升小波算法应用于铁谱图像边缘检测:首先将彩色铁谱图像分解为R、G、B单通道图像,对三通道图像分别进行预处理,并利用直方图处理和图像深度转换实现磨粒和背景的分离;然后对各个通道图像进行D4提升小波变换,在小波域中,通过阈值判断提取高频子图中的边缘像素;最后通过或运算将各个通道的边缘进行融合得到最终的磨粒边缘图像。本文结果与Sobel算子和Canny算子的结果进行比较表明:本文中算法能有效的抑制噪声,较好地再现铁谱图像的磨粒边缘信息,是一种有效的铁谱图像边缘检测算法。
Multiscale wavelet analysis to image edge detection is a current research focus. The edge of grain is the basic feature in the ferrography image, providing important characteristic parameters for ferrography image~ feature extraction and playing a very important role in failure analysis based on ferrographic analysis technology. In this ar- ticle, the lifting wavelet transforms is used to edge detection of ferrograghy image. First, the color image of ferrogra- phy is segmented into RGB single channel images, preconditioning each single image and separating grains from the background by histogram and image depth conversion; Then, IM lifting wavelet is used to each single channel im- age, extracting pixels of edge from the sub-image in wavelet domain by threshold ; Lastly, the final edge images are obtained through OR operation. Compared with the Soble algorithm and the Canny algorithm, this article~ algorithm is more efficient at r^duoln~ n,,;~, t~,,~ ~h. oA~ ~.~：^- -c c L_ _＇___
出处
《机械科学与技术》
CSCD
北大核心
2013年第10期1466-1470,共5页
Mechanical Science and Technology for Aerospace Engineering
基金
国家自然科学基金项目(51205202)资助
关键词
特征提取
故障诊断
提升小波变换
边缘检测
edge detection
grain boundavries
pixels
wavelet analysis
wavelet transforms
algorithms
featureextraction
finite element method
failure analysis
lifting wavelet transforms
