自适应空间约束融入混合模型的遥感图像分割  

Remote Sensing Image Segmentation Based on Mixture Model Infused with Adaptive Spatial Constraint

作　　者：石雪 王玉 SHI Xue;WANG Yu(College of Geomatics and Geoinformation,Guilin University of Technology,Guilin 541004,China)

机构地区：[1]桂林理工大学测绘地理信息学院,广西桂林541004

出　　处：《无线电工程》2023年第1期122-128,共7页Radio Engineering

基　　金：广西自然科学基金(2022GXNSFBA035567,2020GXNSFB A297096)。

摘　　要：为了降低图像噪声的影响并提高遥感图像分割精度,提出了一种自适应空间约束融入混合模型的遥感图像分割算法。考虑到学生t分布具有重尾特性比高斯分布更具有鲁棒性,利用学生t混合模型(Student’s-t Mixture Model, SMM)建模像素光谱测度概率分布。为了避免图像噪声对分割结果的影响,基于马尔可夫随机场利用局部像素类属概率定义组份权重,将像素空间相关性融入SMM,进而构建出空间约束图像分割模型。为了实现自适应平滑系数的模型参数求解,采用梯度下降方法求解分割模型。采用本文算法对添加噪声的遥感图像进行分割实验,结果表明,所提算法可有效降低图像噪声的影响,同时可准确分割高分辨率遥感图像。To reduce the influence of image noise and improve the segmentation accuracy of image, a remote sensing image segmentation algorithm based on a mixture model infused with adaptive spatial constrain is proposed. The Student’s-t distribution has a longer tail and is more robust than Gaussian distribution for image noise. The distribution of pixel spectral measurement is modeled by Student’s-t Mixture Model(SMM). In order to avoid the influence of image noise on segmentation results, the component weight is defined by Markov random field using the attribute probability of local pixels, and the spatial correlation of pixels is integrated into SMM. Then, the spatially constrained image segmentation model is constructed. In order to compute the model parameters of the adaptive smoothing coefficient, the gradient descent method is used to solve the segmentation model. The proposed algorithm is used to segment the remote sensing image with noise. The results show that the proposed algorithm can effectively reduce the infulence of image noise and accurately segment high-resolution remote sensing images.

关 键 词：图像分割 高分辨率遥感图像 学生t混合模型 马尔可夫随机场 自适应空间约束 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置]