基于模糊TS-MRF模型的无监督声纳图像分割

发布时间：2018-06-05 01:56

  本文选题：声纳图像分割 + 树结构化的马尔可夫随机场(TS-MRF)　； 参考：《华中科技大学学报(自然科学版)》2017年05期

【摘要】：为解决声纳图像本身特征信息较弱,而树结构化的马尔可夫随机场(TS-MRF)算法在分割中过分依赖祖先节点,且在标号分割中仅考虑区域内部一致性而忽视区域边缘的各向异性的问题,提出了一种模糊树结构化的马尔可夫随机场(TS-MRF)模型的声纳图像分割算法.在TS-MRF势函数中引入广义模糊算子,以模糊隶属度作为像素相似度度量,将邻域信息融入到分裂节点参数的确定中,使得先验概率的刻画更加精细.已知图像观测特征前提下定义分裂增益系数来反映分裂前、后标号后验概率的比值,并将对增益系数的判断作为确定二叉树节点分裂的依据,降低求解后验概率最大的计算复杂度.结合区域分裂合并方法完成对声纳图像无监督分割.实验结果从视觉效果和客观评价表明:本分割方法相比于传统MRF和TS-MRF等分割算法,具有较高的分割精度和高鲁棒性.
[Abstract]:In order to solve the problem that the feature information of sonar image is weak, the tree structured Markov random field (TS-MRF) algorithm relies too much on the ancestor nodes in segmentation. The anisotropy of the edge of the region is only considered in label segmentation. A fuzzy tree structured Markov random field (TS-MRF) model is proposed for sonar image segmentation. The generalized fuzzy operator is introduced into the TS-MRF potential function and the fuzzy membership degree is used as the pixel similarity measure. The neighborhood information is incorporated into the parameter determination of the split node, which makes the characterization of the priori probability more precise. Based on the known image observation characteristics, the splitting gain coefficient is defined to reflect the ratio of pre-splitting and post-label posteriori probability, and the judgment of gain coefficient is taken as the basis for determining the division of binary tree nodes. Reduce the computational complexity of the maximum posteriori probability. The unsupervised segmentation of sonar image is accomplished by combining region splitting and merging method. The experimental results show that the proposed method has higher segmentation accuracy and robustness than the traditional MRF and TS-MRF segmentation algorithms.
【作者单位】： 三峡大学水电工程智能视觉监测湖北省重点实验室;三峡大学计算机与信息学院;
【基金】：国家自然科学基金(联合基金)重点资助项目(U1401252);国家自然科学基金资助项目(61272237) 湖北省重点实验室开放基金资助项目(2015KLA05)
【分类号】：TP391.41
，

本文编号：1979898