| 基于多源信息与深度学习特征提取的人口空间抽样方法 |
| |
| 引用本文： | 淳锦,张新长,郭海京,张建国,金诗程.基于多源信息与深度学习特征提取的人口空间抽样方法[J].测绘通报,2021,0(8):42-47. |
| |
| 作者姓名： | 淳锦 张新长 郭海京 张建国 金诗程 |
| |
| 作者单位： | 广东省国土资源测绘院,广东 广州510500;广州大学地理科学与遥感学院,广东 广州510006;湖南博通信息股份有限公司,湖南 长沙410007 |
| |
| 基金项目： | 国家自然科学基金面上项目（42071441）；广东省省级科技计划（2018B020207002）；国土空间规划多源数据融合与更新合作研发项目（521023） |
| |
| 摘 要： | 人口抽样调查是通过人口样本估算区域人口总体的一种手段。由于人口分布通常具有空间差异性，传统的抽样调查理论难以满足日益增长的空间抽样需求，合理高效的人口空间抽样调查方法对于人口统计、研究人类活动、解决城市问题等有重要意义。本文提出一种基于多源信息与深度学习特征提取的人口空间抽样方法。在不透水面信息的辅助下，利用四叉树分割进行分层抽样，初步选择出可能存在人口分布的调查样本，并通过深度学习的常用模型——卷积神经网络估算样本建筑物密度，以辅助最终调查样本的选择与调查方案的制定。研究结果证明，该方法能够有效地筛选与人口分布密切相关的抽样区域，排除大量的无用样本，提高了人口调查的效率，节约了大量调查成本。
|
| 关 键 词： | 多源信息 人口抽样 四叉树 深度学习 特征提取 |
| 收稿时间： | 2020-12-10 |
| 修稿时间： | 2021-07-01 |
Spatial sampling of population based on multi-source information and deep learning feature extraction |
| |
| CHUN Jin,ZHANG Xinchang,GUO Haijing,ZHANG Jianguo,JIN Shicheng.Spatial sampling of population based on multi-source information and deep learning feature extraction[J].Bulletin of Surveying and Mapping,2021,0(8):42-47. |
| |
| Authors: | CHUN Jin ZHANG Xinchang GUO Haijing ZHANG Jianguo JIN Shicheng |
| |
| Institution: | 1. Surveying and Mapping Institute Lands and Resource Department of Guangdong Province, Guangzhou 510500, China;2. School of Geographical Sciences and Remote Sensing, Guangzhou University, Guangzhou 510006, China;3. Hunan Botong Information Co., Ltd., Changsha 410007, China |
| |
| Abstract: | Population sampling survey is a means to estimate the population of a region through population samples. Due to the spatial difference of population distribution, the traditional sampling survey theory is difficult to meet the growing demand of spatial sampling. The research on reasonable and efficient population spatial sampling survey method is of great significance to population statistics, human activities and urban problems. This paper proposes a population spatial sampling method based on multi-source information and deep learning feature extraction. Firstly, we use quadtree segmentation for stratified sampling with the help of impervious surface information, initially select the survey samples that may have population distribution.Secondly, we estimate building density of sample by convolution neural network that is a common model of deep learning to assist in the final sample selection and survey scheme formulation. The results show that this method can effectively screen the sampling areas closely related to population distribution, eliminate a large number of useless samples, improve the efficiency of population survey and save a lot of survey costs. |
| |
| Keywords: | multi-source information population sampling quadtree deep learning feature extraction |
| 本文献已被 CNKI 万方数据 等数据库收录！ |
| 点击此处可从《测绘通报》浏览原始摘要信息 |
| 点击此处可从《测绘通报》下载免费的PDF全文 |
|
