面向灌区渠系轮廓提取的DEM处理方法研究

发布时间：2018-04-19 13:19

本文选题：灌区渠系轮廓 + 数字高程模型　；参考：《西北农林科技大学》2017年硕士论文

【摘要】：灌区渠系的快速制图对节水灌溉、合理配水以及安全输水等具有重要意义。针对目前灌区渠系轮廓提取过程中存在的遥感影像识别率低、提取较难以及人工干预较多等问题,本文以无人机航空摄影测量获得的正射影像、数字高程模型(Digital Elevation Model,DEM)和坡度数据为基础,分别采用了面向对象分类方法、监督分类方法、基于地形特征方法和改进的霍夫变换方法分别实现渠系轮廓提取。由于坡度数据能够减少灌区非渠系因素中道路以及农地对渠系网络的影响,使得渠系特征得到增强,因此,与手动数字化结果对比,改进的霍夫变换方法在坡度数据上的提取完整度最高,优于面向对象分类法、监督分类法以及基于地形特征在多种数据上的渠系轮廓提取结果。该方法是对河套灌区渠系快速制图方法的一次有效探索。主要研究内容及结果如下:(1)设计并实现了基于地形特征的渠系轮廓提取方法。首先将灌区中的公路、湖泊及房屋等干扰因素作预处理,进行过滤,并对DEM进行坡度转换,消除几乎无坡度变化的田间道路数据;其次通过对数据集进行测试,查找区分渠系与非渠系信息的最优阈值,得到区域分割结果;最后根据实地勘察过程中的采样点信息,按照广度优先搜索方式进行渠系轮廓信息的搜寻及保存。经证明,基于地形特征的渠系轮廓提取方法完整度为66.95%,高于面向对象方法在多种数据上的渠系轮廓提取结果,低于监督分类方法在多种数据上的渠系轮廓提取结果,错误率最低,仅为34.72%。(2)设计并实现了改进霍夫变换的渠系轮廓提取方法。在灌区中,渠系和田间农地、道路具有规整及统一的线性结构,而田间农地和道路基本无坡度变化,因此坡度转化过程能够剔除掉与渠系具有相似分布的道路和农地信息。本研究以坡度数据为基础,采用灰度化、滤波去噪、二值化以及连通区域干扰点去除等步骤,结合基本霍夫变换能够在较多噪声环境下提取直线段的特点,实现了对灌区渠系轮廓的获取,并对结果中出现的渠系断开以及相交问题使用几何方法进行有效改进。经证明,改进霍夫变换的渠系轮廓提取完整度最高,可达到85.61%,比面向对象分类法坡度结果高出26.91%,比该方法的错误率低47.33%;比监督分类方法坡度结果高出3.77%,但是错误率低23.34%。
[Abstract]:Rapid mapping of canal system in irrigation area is of great significance to water saving irrigation, reasonable water distribution and safe water transportation.Aiming at the problems of low recognition rate, difficult extraction and more manual intervention in the process of contour extraction of canal system in irrigation area, the orthophoto image obtained by aerial photogrammetry of UAV is used in this paper.Based on the digital elevation model (DSM) and slope data, object oriented classification method, supervised classification method, terrain feature method and improved Hough transform method are used to extract canal system contour, respectively.Because slope data can reduce the influence of road and farmland on canal system network in irrigation area, the characteristics of canal system are enhanced, so compared with manual digitization results, the slope data can reduce the influence of road and farmland on canal system network.The improved Hough transform method has the highest extraction integrity on slope data, which is superior to the object-oriented classification method, the supervised classification method and the contour extraction results of canal system based on terrain features on a variety of data.This method is an effective exploration for rapid mapping of canal system in Hetao irrigation area.The main research contents and results are as follows: (1) the method of canal system contour extraction based on terrain feature is designed and implemented.Firstly, the disturbance factors such as roads, lakes and houses in irrigated area are pretreated, filtered, and the slope degree of DEM is converted to eliminate the field road data with almost no slope change; secondly, the data set is tested.Find the optimal threshold to distinguish the canal system information from the non-canal system information and get the region segmentation result. Finally according to the sampling point information in the field survey the contour information of canal system is searched and preserved according to the breadth-first search method.It has been proved that the integrity degree of canal system contour extraction method based on terrain feature is 66.95, which is higher than that of object-oriented method on many kinds of data, and lower than that of supervised classification method on many kinds of data.The error rate is the lowest, only 34.72%.) the improved Hough transform is designed and implemented to extract the contour of canal system.In irrigation area, the road has a regular and uniform linear structure in canal system and field farmland, but there is no slope change in field farmland and road, so the information of road and farmland with similar distribution can be eliminated in the process of slope transformation.Based on the slope data, this study adopts the steps of grayscale, filtering and denoising, binarization and the removal of interference points in the connected region, combining with the characteristics of the basic Hough transform to extract the straight line segment in the environment of more noise.The contour of canal system in irrigation area is obtained, and the geometric method is used to improve the problem of disconnection and intersection of canal system in the result.It has been proved that the improved Hough transform has the highest degree of integrity of canal system contour extraction, which can reach 85.61, 26.91 higher than the result of object-oriented classification, 47.33 lower than the error rate of this method, and 3.77 higher than that of supervised classification method, but the error rate is 23.34 points lower.
【学位授予单位】：西北农林科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：S27;TP79

【参考文献】