地理国情监测数据自动变化检测技术研究及系统研发

发布时间：2018-06-11 18:33

本文选题：地理国情监测 + 影像分割　；参考：《西南交通大学》2017年硕士论文

【摘要】：随着经济高速、稳步的发展,我国地表自然和人文信息的变化更趋频繁。为进一步满足政府、社会和大众对土地利用变化信息实时性掌握的需求,我国全面开展了地理国情监测。传统地理国情监测方法工作量大、自动化程度不高、效率低下,从而限制了地理国情数据的更新周期的缩短。相反,遥感技术作为一种非接触的探测技术,具有探测范围广、获取地表真实信息速度快、周期短等特点,可对地理国情进行快速、大范围的监测,从而满足地理国情对更新周期的需要。但是,由于传统的遥感技术变化检测需要两幅或两幅以上遥感影像,而常态化地理国情监测中往往只能获取前期矢量数据和后期遥感影像。因此,为解决这一矢量数据预更新问题,本文在现有遥感影像变化检测技术上对矢量数据和遥感影像联合变化检测方法做进一步研究和探讨。本文的主要研究内容如下:(1)对基于矢量数据与遥感影像的变化检测技术的概念及流程进行了阐述,并按照在检测过程中是否需要选取样本将其分为监督法和非监督法。对影像分割、特征提取、阈值自动获取等矢量与遥感影像变化检测中的三个关键技术进行了归纳和总结。(2)研究了基于像斑类别异质度的矢量数据与遥感影像的变化检测方法。该方法借鉴了经典的标记分水岭分割方法,以矢量数据作为先验边界,对影像进行约束分割,获取同质像斑;提取像斑特征,构建像斑类别异质度,通过自动化阈值获取技术获取每一地物类别异质度阈值,最后进行变化/未变化像斑的判别。该方法实现了矢量数据与遥感影像的自动变化检测。(3)研发了地理国情常态化监测数据变化检测系统。以基于像斑类别异质度的矢量数据与遥感影像变化检测方法为技术基础,采用GDAL(Geospatial Data Abstraction Library)与ArcEngine混合编程技术建设了地理国情常态化监测数据变化检测系统。系统采用C#语言,在Microsoft Visual Studio 2010平台上,利用GDAL对栅格数据进行处理,利用ArcEngine对矢量数据进行处理。系统实现了栅格影像与矢量数据的叠加显示、矢量数据约束下的影像分割、像斑特征提取、特征距离度量、阈值获取、变化检测、人工编辑(新增、删除、修改)、参数设定(分割参数)、数据输出(分割结果、变化检测结果)、后期编辑等功能。该系统能够实现自动化的变化检测,与传统的目视解译相比,变化检测效率得到较大的提高。(4)以新都区某村2015年地理国情普查矢量数据的地表覆盖层与2016年航空高分遥感影像进行变化检测,变化检测的正确率、虚检率、漏检率分别为84.3%、、15.6%、16.1%,变化检测结果验证了该方法的可行性与有效性,变化检测精度已达到地理国情监测中的精度要求。该变化检测同时证明了研发的系统可有效的辅助地理国情数据中变化区域的发现,提高内业更新的效率。
[Abstract]:With the rapid and steady development of economy, the changes of natural and humanistic information on the surface of our country become more frequent. In order to further meet the needs of the government, society and the public to grasp the real-time information of land use change, our country has carried out the monitoring of geographical conditions in an all-round way. The traditional method of geographical situation monitoring has the advantages of heavy workload, low automation and low efficiency, which limits the shortening of the updating period of geographical national data. On the contrary, as a non-contact detection technology, remote sensing technology has the characteristics of wide range of detection, fast speed of obtaining real information on the surface, short period and so on. In order to meet the geographical conditions of the update cycle needs. However, two or more remote sensing images are needed for the change detection of traditional remote sensing technology. However, in the normal geographical situation monitoring, only the former vector data and the later remote sensing images can be obtained. Therefore, in order to solve the problem of vector data pre-updating, this paper further studies and discusses the joint change detection method of vector data and remote sensing image on the existing remote sensing image change detection technology. The main contents of this paper are as follows: (1) the concept and flow of change detection technology based on vector data and remote sensing image are expounded, and they are divided into supervised method and unsupervised method according to whether it is necessary to select samples in the process of detection. In this paper, three key techniques of image segmentation, feature extraction, threshold automatic acquisition and remote sensing image change detection are summarized and summarized. (2) the change detection method of vector data and remote sensing image based on the heterogeneity of image spot category is studied. This method uses the classical marking watershed segmentation method for reference, uses vector data as the prior boundary, performs constrained segmentation on the image, obtains the homogeneous image spot, extracts the image spot feature, and constructs the heterogeneity degree of the image spot category. The heterogeneity threshold of each feature category is obtained by the automatic threshold acquisition technique. Finally, the variable / unchanged image spots are identified. This method realizes the automatic change detection of vector data and remote sensing image. Based on the change detection method of vector data and remote sensing image based on the heterogeneity of image spot category, the change detection system of the normal monitoring data of geographical national conditions is built by using the mixed programming technology of GDAL spatial data Abstraction library and ArcEngine. The system uses C # language, on the platform of Microsoft Visual Studio 2010, uses GDAL to process raster data and ArcEngine to process vector data. The system realizes the superposition display of raster image and vector data, image segmentation under vector data constraint, image spot feature extraction, feature distance measurement, threshold value acquisition, change detection, manual editing (add, delete, delete). Modification, parameter setting (partition parameters, data output (segmentation results, change detection results, late editing and so on). Compared with traditional visual interpretation, the system can realize automatic change detection. The efficiency of change detection has been greatly improved.) change detection is carried out on the ground overlay of a village's 2015 geographic situation census vector data and 2016 aerial high score remote sensing image. The accuracy and false detection rate of change detection are obtained. The rate of missing detection is 84.3and 15.6and 16.1respectively. The result of change detection verifies the feasibility and validity of the method, and the accuracy of the change detection has reached the precision requirement of geographical situation monitoring. The change detection also proves that the developed system can effectively assist the discovery of the changing regions in the geographical national conditions data and improve the efficiency of the renewal of the internal industry.
【学位授予单位】：西南交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：P237

【参考文献】