论文部分内容阅读
针对噪声导致高分辨遥感影像分割存在过分割或者欠分割的问题,提出结合相位一致和分水岭变换的高分辨率影像分割方法。该方法首先采用基于光谱相似性的相位一致的模型方法来获得边缘响应幅度,再采用自动标记分水岭算法对影像进行分割;基于相邻分割对象的空间位置、形状、面积等特征多重约束,提出相邻分割对象合并代价函数模型,对分割结果进行优化并获取最终分割结果。选择典型地区实验影像进行分割实验,通过目视评价和监督评价,并与典型分割方法进行比较,验证所提分割方法的有效性。
Aiming at the problem of high resolution remote sensing image segmentation caused by noise, the high resolution image segmentation method is proposed based on the consistent phase and watershed transform. Firstly, a phase-consistent model based on spectral similarity was used to obtain the amplitude of the edge response. Then, an automatic labeling watershed algorithm was used to segment the image. Based on the feature multiple constraints such as spatial location, shape and area of adjacent segmentation objects, The neighbor segmentation object merges the cost function model, optimizes the segmentation result and obtains the final segmentation result. The experimental images of typical regions were selected for segmentation experiment. The visual evaluation and supervisory evaluation were compared with the typical segmentation methods to verify the validity of the proposed segmentation method.