基于改进多尺度形态学的裂缝图像去噪算法  

DENOSING ALGORITHM OF CRACK IMAGE BASED ON IMPROVED MULTISCALE MORPHOLOGY

作　　者：徐劲力[1] 李征瑞 黄丰云 许建宁 Xu Jinli;Li Zhengrui;Huang Fengyun;Xu Jianning(School of Mechanical and Electrical Engineering,Wuhan University of Technology,Wuhan 430070,Hubei,China)

机构地区：[1]武汉理工大学机电工程学院,湖北武汉430070

出　　处：《计算机应用与软件》2023年第3期255-259,265,共6页Computer Applications and Software

基　　金：中央高校基本科研业务费项目(205204004)。

摘　　要：AGV小车行驶场地中地面的光照不均、对比度低,背景复杂会导致裂缝图片采集时图像偏暗并带有较多混合噪声,传统去噪算法不能有效去噪。针对传统算法的不足,提出一种基于改进多尺度形态学的裂缝图像去噪算法。选取一个合适的结构元素,将其循环膨胀得到形状相同大小不同的多个结构元素,用得到的多个结构元素依次对裂缝图像开运算处理并将处理结果叠加,再使用纹理均衡的补偿方法处理结果并输出。通过与传统中值滤波、均值滤波、多尺度形态学滤波等方法比较,提出的新方法不仅能够降低图像中的噪声,有效地保护裂缝边缘,同时有利于裂缝特征的分割。During the acquisition of the crack image,the uneven illumination and low contrast of the AGV driving field and the complex background will make the crack image darker and increase the mixed noise content.The traditional denoising algorithm cannot effectively denoise.In view of the shortcomings of traditional algorithms,a crack image denoising algorithm based on improved multiscale morphology is proposed.A suitable structural element was selected,expanding it cyclically to obtain multiple structural elements with the same shape and different size.The morphological opening was operated to eliminate the crack image with multiple structural elements,superimposing the output.The image by texture equalization was processed and the results were output.Compared with traditional median filtering,mean filtering and multiscale morphological filtering methods,the proposed new method can not only reduce the noise in the image,effectively protect the crack edges,but also is conducive to cracks feature segmentation.

关 键 词：多尺度形态学 纹理均衡 裂缝 去噪 

分 类 号：TP3[自动化与计算机技术—计算机科学与技术] TP751.1