高分辨率遥感影像林地边界提取方法

由于高分辨率遥感影像上的信息高度细节化,加之噪声的影响,会导致基于像元级纹理特征的林地边界提取方法的效果不理想。为此,提出一种基于种子纹理基元合并的半自动林地边界提取方法。首先利用基于图模型的影像分割算法获取初始基元;然后定义了一种针对非规则基元统计基元级灰度共生矩阵(GLCM)纹理特征的方法;最后在人工给定种子基元的基础上合并具有相似纹理的基元,并对基元合并的结果进行边界提取,得到高分影像上的林地边界。利用多源高分影像对所提方法进行验证及对比分析。实验结果表明,该方法对高分影像上大片典型林地的边界可取得较高的提取精度和计算效率。

Forest Boundary Extraction Method Applied to High Resolution Remote Sensing Images

Due to the highly detailed information and noise in high resolution remote sensing images, the results of traditional forest boundary extraction algorithms based on pixel-based texture features are not satisfactory. There-fore, a semi-automatic method based on seeded texture primitive merging was proposed in this paper. Firstly, the initial primitives were obtained by the graph-based image segmentation algorithm, and then a new primitive-based Gray Level Co-occurrence Matrix ( GLCM) texture feature extraction method was defined and directly applied to the irregular primitives. Based on the seed primitives provided artificially, the proposed algorithm merged primitives with similar texture features and applied a boundary extraction algorithm to the result of texture primitive merging in order to extract the forest boundary. In the experiment, multi-source high resolution remote sensing images were used to validate the proposed method. The comparative analysis with other methods shows that the proposed method can extract the boundary of large typical forests from high resolution remote sensing images with a higher extraction accuracy and computational efficiency.