基于并行GVF Snake模型的黄土地貌沟沿线提取  

Extraction of loess landform shoulder line based on parallel GVF Snake model

作　　者：宋效东[1] 汤国安[1] 周毅[1] 田剑[1,2] 

机构地区：[1]南京师范大学虚拟地理环境教育部重点实验室,江苏南京210046 [2]合肥工业大学资源与环境工程学院,安徽合肥230009

出　　处：《中国矿业大学学报》2013年第1期134-140,共7页Journal of China University of Mining & Technology

基　　金：国家高技术研究发展计划(863)项目(2011AA120303);国家自然科学基金项目(40930531;41171299);江苏省普通高校研究生科研创新计划项目(CXZZ12_0393)

摘　　要：提出了一种改进的并行GVF Snake算法,用来提取大范围高分辨率黄土高原地区的沟沿线.该算法使用规则格网数字高程模型数据,基于全局梯度向量场提取各计算节点的沟沿线.结合沟沿线特殊的空间位置,提出初始轮廓线自动设定的方法.通过改善初始轮廓的自动设定,大大提高了沟沿线提取的准确性,同时也降低了GVF Snake模型的计算时间.在9节点的机群系统上对算法的性能和实验结果准确性进行了测试.在陕北黄土高原梁峁丘陵沟壑区的实验结果表明,本算法可准确地将初始轮廓线设置在有效逼近域内,大大提高了抗干扰性,能够实现黄土地貌沟沿线准确、有效的自动提取;同时,也可获得良好的并行加速比,并行效率较高.Parallel gradient vector flow （GVF） Snake model is introduced in this paper to extract the wide-range and high-resolution shoulder line in loess plateau. This model uses the regular square grid data, and extracts the shoulder line of each compute node based on global gradient vector field. Then, combined with spatial positions of shoulder line, an automatic algorithm is proposed to select the initial contour line. With the consideration of improving the auto-setting of initial contour, the accuracy of extracting shoulder line is also greatly improved, and computing time of GVF Snake model is reduced at the same time. The performance of computing and accuracy of experimental results are also evaluated on a cluster with 9 nodes. By testing Liangmao hilly-gully region of loess landform in Shaanxi Province, experimental results show that this kind of computing can accurately set the initial contour in the effective approaching domain, and improve the anti-interference. This can realize the goal of accurately and effectively auto-exaction of should line in loess landform, and obtain better parallel speed-up ratio and high-parallel effectivenss.

关 键 词：黄土地貌 沟沿线 GVF SNAKE 并行计算 DEM 

分 类 号：P208[天文地球—地图制图学与地理信息工程]