一种基于自适应M-S模型的遥感影像分割方法

A High Resolution Remote Sensing Image Segmentation Method Based on Adaptive M-S Model

提出一种基于自适应M-S模型的遥感影像多特征融合的分割方法。首先结合改进的Sobel算子进行阈值化轮廓提取方法提取边缘信息;然后利用波段距离加权函数计算边权值,同时按照一定的原则加入边缘特征,采用最小生成树算法获得初始分割对象;最后在光谱特征和纹理特征的辅助下进行自适应M-S模型合并,合并后的对象即为分割结果。为了验证该方法的可行性,采用Quickbird影像和高分二号影像进行实验分析并对结果做出定性和定量评估。实验结果表明,基于自适应M-S模型的遥感影像分割方法的分割精度优于分形网络演化算法,同时分割速度也略有提升。

A segmentation method for multi-feature fusion of remote sensing images is proposed based on adaptive Mumford-Shah model. Firstly, the improved Sobel operator is used to extract the edge information from the threshold contour extraction method. Then, the edge weight is calculated using the band distance weighting function, and the edge features are added in accordance with certain principles. The minimum spanning tree algorithm is used to obtain the initial segmentation objects. The final segmentation objects are produced with the aid of spectral information and texture information in accordance with the adaptive Mumford-Shah model(M-S model). In order to verify the feasibility of the method, Quickbird images and GF-2 images are used to carry out experimental analysis, qualitative and quantitative evaluation of the results. The results show that the segmentation precision of the proposed algorithm is better than that of the fractal network evolution algorithm, and the segmentation speed is also slightly improved.