由于遥感图像景观边缘具有复杂非线性结构，受光照与场景影响，使得图像边缘存在极多样化纹理和噪声，导致景观边缘难以捕捉。于是提出基于超像素分割算法的景观边缘提取方法。采用超像素分割算法，将图像映射至CIE-LAB(Inter-national Commission on Illumination-Lightness-L∗A∗B)色彩空间中，利用最小化梯度，将超像素置于梯度对应区域中，获取标准化度量，将水平与垂直梯度加入特征矢量中，组建景观弱边缘超像素分割模型，分割景观弱边缘超像素，利用高斯滤波器，模糊处理超像素分割结果，构建能量函数，通过迭代处理，实现景观边缘提取。实验结果表明，所提方法能准确分割显著边缘与细节边缘，精准提取景观边缘，且边缘线连贯、平滑，有效避免了伪轮廓问题，对实现城市与景观的合理规划具有较高的实际应用价值。
Simulation of Landscape Edge Extraction Based on Superpixel Segmentation Algorithm
Due to the complex nonlinear structure of landscape edges in remote sensing images,which are affected by lighting and scene,there are extremely diverse textures and noise at the edges of the image,making it difficult to capture the landscape edges.Therefore,a method of extracting landscape edges based on superpixel segmentation al-gorithm was proposed.Firstly,the image was mapped to International Commission on Illumination-Lightness-L∗A∗B(CIE-LAB)color space by the super-pixel segmentation algorithm.After the gradient was minimized,the super-pixel was placed in the gradient region correspondingly,thus obtaining standardized measurements.Then,the horizontal and vertical gradients were added into the feature vector to build a model of weak edge superpixel segmen-tation model.Moreover,Gaussian filter was used to blur the superpixel segmentation results and establish energy func-tions.Finally,the landscape edge extraction was achieved by iterative processing method.The experimental results show that the proposed method can accurately segment significant edge and detailed edge,and extract landscape edge.Meanwhile,the edge lines are coherent and smooth,thus avoiding the pseudo-contour problem.For this reason,the method has high practical application value for rational planning of city and landscape.
NETL
NSTL
万方数据
