邻域约束高斯混合模型的模糊聚类图像分割

发布时间：2018-03-08 16:03

本文选题：高斯混合模型(GMM)　切入点：邻域约束　出处：《模式识别与人工智能》2017年03期 　论文类型：期刊论文

【摘要】：针对传统模糊聚类分割方法无法有效模拟数据分布特征的问题,提出基于邻域约束高斯混合模型的模糊聚类图像分割算法.利用高斯分布刻画聚类内像素光谱测度统计特征,定义像素与其邻域像素相关性的先验概率,并作为高斯混合模型中各高斯分量权重系数,构建包含特征场邻域作用的高斯混合模型.利用高斯分量描述像素与聚类间的非相似性测度,建立基于高斯混合模型的模糊聚类目标函数.在传统模糊聚类方法基础上,采用高斯混合模型定义像素与聚类间的非相似性测度,并在高斯混合模型中融入邻域作用,有效解决数据具有多峰值特征的问题.最后通过实验验证文中算法的准确性.
[Abstract]:Aiming at the problem that the traditional fuzzy clustering segmentation method can not effectively simulate the characteristics of data distribution, a fuzzy clustering image segmentation algorithm based on neighborhood constrained Gao Si hybrid model is proposed. A priori probability of the correlation between pixels and their neighboring pixels is defined, and used as the weight coefficients of each Gao Si component in Gao Si's mixed model. In this paper, a Gao Si hybrid model with the neighborhood effect of feature field is constructed. Then, the objective function of fuzzy clustering is established by using Gao Si component to describe the dissimilarity between pixels and clusters. Based on the traditional fuzzy clustering method, the objective function of fuzzy clustering is established. Gao Si's mixed model is used to define the dissimilarity measure between pixels and clustering, and the neighborhood function is incorporated into Gao Si's mixed model, which effectively solves the problem of multi-peak data. Finally, the accuracy of the proposed algorithm is verified by experiments.
【作者单位】：辽宁工程技术大学测绘与地理科学学院遥感科学与应用研究院;
【基金】：国家自然科学基金面上项目(No.41271435);国家自然科学基金青年基金项目(No.41301479) 辽宁省自然科学基金项目(No.2015020190)资助~~
【分类号】：TP391.41

【相似文献】