基于气泡稳定性分析的锌浮选泡沫图像时空联合去噪
Stability characteristics-based zinc-flotation froth image denoising fusing spatial-temporal information
摘要点击 56  全文点击 47  投稿时间:2018-12-04  修订日期:2019-07-05
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2019.80956
  2020,37(4):721-730
中文关键词  浮选泡沫图像  气泡稳定性分析;非局部均值;时空联合去噪
英文关键词  flotation froth image  bubble stability analysis  nonlocal mean  spatial-temporal joint denoising
基金项目  国家自然科学基金(61771492,61501183);国家自然科学基金-广东联合基金(U1701261).国家杰出青年科学基金(61725306)
学科分类代码  
作者单位E-mail
肖文辉 中南大学 信息科学与工程学院 xiaowenhui@csu.edu.cn 
唐朝晖 中南大学 信息科学与工程学院  
刘金平 湖南师范大学 信息科学与工程学院 ljp202518@163.com 
谢永芳 中南大学 信息科学与工程学院  
中文摘要
      浮选泡沫运动过程中不可避免地出现形变、坍塌、兼并、破裂等动态变化特性,常用的去噪方法难以获得高质量的监测图像,提出一种基于气泡稳定性分析的泡沫图像时空联合去噪方法。该方法采用扩展的相位相关法对浮选气泡进行亚像素运动估计,通过双线性插值进行运动补偿;在此基础上,以泡沫图像子块为单位检测气泡的稳定性,准确辨识出泡沫图像子块的稳定运动状态(Stable motion state, SMS)和非稳定运动状态(Unstable motion state, UMS);对具有SMS特性的子块采用时域滤波去噪,对具有UMS特性的子块采用非局部均值(Nonlocal means,NLM)方法进行空域滤波去噪;并根据气泡子块的相关系数,联合时域滤波结果和空域滤波结果获得泡沫图像的时空联合去噪输出。在锌浮选过程监控中进行实验验证,结果表明,该方法可以获得高信噪比的泡沫图像,去噪结果结构相似性强,为泡沫视觉特征的准确提取奠定了基础。
英文摘要
      During the process of flotation froth movement, deformation, collapse, annexation, rupture and other dynamic characteristics inevitably occur, so it is difficult to obtain high quality monitoring image by common de-noising methods. A spatial-temporal joint denoising method for froth images based on bubble stability analysis is proposed. The extended phase correlation method is used to estimate the subpixel motion of the flotation froth and the motion compensation is carried out by bilinear interpolation. On this basis, the bubble stability is detected by using the froth image sub-block as the unit, and the stable motion state (SMS) and the unstable motion state (UMS) of the froth image sub-block are identified accurately. The sub-blocks with SMS characteristics are de-noised by time-domain filtering, and the sub-blocks with UMS characteristics are de-noised by non-local mean (NLM) method in spatial domain. According to the correlation coefficient of bubble sub-block, combined time domain filtering and spatial filtering, the spatial-temporal joint denoising output of froth image is obtained. The experimental results in zinc flotation process show that this method can obtain froth images with high PSNR, and the denoising results have strong structural similarity, which lays a foundation for accurate extraction of froth visual features.