启发法优化的农田地区高分遥感影像分割

发布时间：2018-08-06 21:58

【摘要】：目前主流的高分遥感影像(high resolution remote sensing image,HRI)分割算法在区域合并顺序的确定中很少考虑区域自身的分割质量信息。针对该问题,提出了一种启发法优化的策略,以提高农田地区HRI的分割精度。首先,提出了区域内和区域间的"均一致模型",前者是利用区域内光谱变化信息来建模的,后者则综合考虑了区域间多光谱与植被信息提取的边界强度;其次,将区域内和区域间的均一致模型合并,构建启发法的执行标准;最后,利用该标准使区域合并的搜索过程能够考虑待合并区域自身的分割质量,从而有效抑制过分割与亚分割错误。利用2景不同特点的农田地区HRI进行分割实验,并将所提出的启发法与2种新提出的分割算法进行对比分析。对分割结果的定量评价结果表明,启发法优化策略可以显著提高农田地区HRI的分割精度。
[Abstract]:At present, the mainstream segmentation algorithms of high score remote sensing image (high resolution remote sensing imageHRI) seldom consider the quality information of the region itself in determining the sequence of region merging. To solve this problem, a heuristic optimization strategy is proposed to improve the precision of HRI segmentation in farmland. First of all, a "uniform model" is proposed, in which the former model is based on the spectral variation information in the region, the latter takes into account the boundary strength of the multi-spectral and vegetation information extraction. The uniform model within and between regions is combined to construct the heuristic standard. Finally, the search process of region merging can consider the segmentation quality of the region to be merged. In order to effectively suppress over-segmentation and sub-segmentation errors. In this paper, the experiment of HRI segmentation in farmland with two different features is carried out, and the proposed heuristic method is compared with two new segmentation algorithms. The quantitative evaluation results of the segmentation results show that the heuristic optimization strategy can significantly improve the segmentation accuracy of HRI in farmland areas.
【作者单位】：内蒙古农业大学水利与土木建筑工程学院;
【基金】：国家自然科学基金项目“科尔沁沙地典型生态系统水热通量传输机理及其与植被耦合关系试验和模拟研究”(编号:51569017);“面向对象的河套灌区遥感作物分类算法研究”(编号:61701265) 内蒙古自然科学基金项目“半干旱区沙地典型生态系统水热通量传输机理研究”(编号:2015MS0514) 中国博士后科学基金面上资助项目“西部地区博士后人才资助计划”(编号:2015M572630XB)共同资助
【分类号】：TP751

【相似文献】