基于图模型和流形排序的遥感影像人工地物提取
发布时间：2019-01-21 15:10
【摘要】：人工地物自动提取是目标识别领域的重要研究内容。该文将图模型和流形排序算法相结合,提出一种基于遥感影像自动提取人工地物的算法。该算法在基于量子遗传算法的KSW熵分割算法获得影像人工地物先验值的基础上,构建以色调、纹理、方向等多特征为邻接边权重的图模型,并以超像素作为基本单位在图模型中进行流形排序,最终通过自适应阈值的方式获得人工地物分割结果。通过利用两幅不同分辨率遥感影像对该算法进行验证,结果显示该算法可以较为完整地提取人工地物,具有较高的正确识别率和较低的漏检率。
[Abstract]:Automatic extraction of artificial objects is an important research content in the field of target recognition. This paper presents an algorithm for automatic extraction of artificial objects based on remote sensing images by combining graph model with manifold sorting algorithm. Based on the KSW entropy segmentation algorithm based on quantum genetic algorithm (QGA) to obtain the priori values of artificial ground objects, a graph model with color, texture, direction and other features as adjacent edge weights is constructed. The superpixel is taken as the basic unit to sort the manifold in the graph model, and finally the artificial ground object segmentation results are obtained by adaptive threshold method. Two remote sensing images with different resolution are used to verify the algorithm. The results show that the algorithm can extract artificial objects completely and has higher correct recognition rate and lower missed detection rate.
【作者单位】： 武汉大学遥感信息工程学院;
【基金】：广东省省级科技计划项目(2015A090905002) 广州市产学研协同创新民生科技重大专项项目(201508020054) 国家863计划项目(2013AA063905)
【分类号】：P237
本文编号：2412768
[Abstract]:Automatic extraction of artificial objects is an important research content in the field of target recognition. This paper presents an algorithm for automatic extraction of artificial objects based on remote sensing images by combining graph model with manifold sorting algorithm. Based on the KSW entropy segmentation algorithm based on quantum genetic algorithm (QGA) to obtain the priori values of artificial ground objects, a graph model with color, texture, direction and other features as adjacent edge weights is constructed. The superpixel is taken as the basic unit to sort the manifold in the graph model, and finally the artificial ground object segmentation results are obtained by adaptive threshold method. Two remote sensing images with different resolution are used to verify the algorithm. The results show that the algorithm can extract artificial objects completely and has higher correct recognition rate and lower missed detection rate.
【作者单位】： 武汉大学遥感信息工程学院;
【基金】：广东省省级科技计划项目(2015A090905002) 广州市产学研协同创新民生科技重大专项项目(201508020054) 国家863计划项目(2013AA063905)
【分类号】：P237
【相似文献】
相关期刊论文 前4条
1 刘亚文,叶晓新;城区人工地物变化检测方法的研究[J];测绘通报;2001年07期
2 刘建贵,张兵,郑兰芬,童庆禧;基于光谱特征的城市人工地物分级分类方法研究[J];遥感技术与应用;1999年03期
3 李二森;张保明;郭海涛;何乔;初艳锋;卢俊;;两种单张航片确定人工地物比高方法的比较[J];测绘科学;2008年02期
4 ;[J];;年期
相关硕士学位论文 前2条
1 李二森;基于遥感影像的人工地物高度信息快速提取技术研究[D];解放军信息工程大学;2008年
2 李英芳;QUICKBIRD影像人工地物提取方法研究[D];中南大学;2010年
，本文编号：2412768
本文链接：https://www.wllwen.com/kejilunwen/dizhicehuilunwen/2412768.html
