融合空间关系的遥感图像分类
发布时间：2018-09-03 06:28
【摘要】：针对光谱纹理特征分类方法的不足,提出了一种融合空间关系的遥感图像分类方法。利用直方图提取像斑特征,采用G统计量构建单像斑概率,通过迭代统计方法计算地物类别邻接概率,利用地物类别邻接概率表达像斑邻域概率,加权组合单像斑概率与像斑邻域概率构建像斑联合概率,依据最大后验概率准则获取图像分类结果。在QucikBird图像上的试验结果表明:与传统的光谱纹理分类方法相比,该方法能够提高图像分类的精度;整体分类精度与Kappa系数分别提高了1.5%和2.1%。
[Abstract]:Aiming at the shortage of spectral texture feature classification method, a remote sensing image classification method combining spatial relationship is proposed. Using histogram to extract the feature of image spot, using G statistic to construct the probability of single image spot, calculating the probability of object category contiguity by iterative statistical method, and using the probability of object class adjacency to express the probability of image spot neighborhood. The joint probability of image spot is constructed by weighted combination single spot probability and image spot neighborhood probability, and the result of image classification is obtained according to the maximum a posteriori probability criterion. The experimental results on QucikBird images show that compared with the traditional spectral texture classification method, this method can improve the accuracy of image classification, the overall classification accuracy and the Kappa coefficient are improved by 1.5% and 2.1%, respectively.
【作者单位】： 四川省第三测绘工程院;
【基金】：测绘地理信息公益性行业科研专项“卫星遥感与地面传感网一体化的湖泊流域地理国情监测关键技术研究”(编号:201512026) 数字制图与国土信息应用工程国家测绘地理信息局重点实验室开放基金资助项目“基于遥感影像的矢量图更新关键技术研究”(编号:DM2016SC04)共同资助
【分类号】：TP751
[Abstract]:Aiming at the shortage of spectral texture feature classification method, a remote sensing image classification method combining spatial relationship is proposed. Using histogram to extract the feature of image spot, using G statistic to construct the probability of single image spot, calculating the probability of object category contiguity by iterative statistical method, and using the probability of object class adjacency to express the probability of image spot neighborhood. The joint probability of image spot is constructed by weighted combination single spot probability and image spot neighborhood probability, and the result of image classification is obtained according to the maximum a posteriori probability criterion. The experimental results on QucikBird images show that compared with the traditional spectral texture classification method, this method can improve the accuracy of image classification, the overall classification accuracy and the Kappa coefficient are improved by 1.5% and 2.1%, respectively.
【作者单位】： 四川省第三测绘工程院;
【基金】：测绘地理信息公益性行业科研专项“卫星遥感与地面传感网一体化的湖泊流域地理国情监测关键技术研究”(编号:201512026) 数字制图与国土信息应用工程国家测绘地理信息局重点实验室开放基金资助项目“基于遥感影像的矢量图更新关键技术研究”(编号:DM2016SC04)共同资助
【分类号】：TP751
【相似文献】
相关期刊论文 前10条
1 黄宁,刘小军,朱敏慧,张守融;遥感图像分类技术研究[J];测试技术学报;2001年02期
2 刘庆生,刘高焕,蔺启忠,王志刚;基于逻辑斯蒂模型的遥感图像分类[J];国土资源遥感;2001年01期
3 谭衢霖,邵芸;雷达遥感图像分类新技术发展研究[J];国土资源遥感;2001年03期
4 杜凤兰,田庆久,夏学齐;遥感图像分类方法评析与展望[J];遥感技术与应用;2004年06期
5 李石华,王金亮,毕艳,陈姚,朱妙园,杨帅,朱佳;遥感图像分类方法研究综述[J];国土资源遥感;2005年02期
6 付小勇;杨建祥;谭靖;;基于统计的遥感图像分类方法[J];林业调查规划;2005年06期
7 王一达;沈熙玲;谢炯;;遥感图像分类方法综述[J];遥感信息;2006年05期
8 李华;曹卫彬;刘姣娣;;土地监测中提高遥感图像分类精度的方法研究[J];安徽农学通报;2008年22期
9 岳昔娟;张勇;黄国满;;改进的直方图均衡化在遥感图像分类中的应用[J];四川测绘;2008年04期
10 曾联明;吴湘滨;刘鹏;;感兴趣区域遥感图像分类与支持向量机应用研究[J];计算机工程与应用;2009年06期
相关会议论文 前3条
1 周军其;张红;孙家b，
本文编号：2219178
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2219178.html
