海岛海岸带地物提取及动态变化分析研究

发布时间：2018-07-31 09:08

【摘要】：海岛海岸带区域资源丰富、经济发达,由于海域开发、城市扩展严重导致海岛海岸带土地利用变化较快。因此,及时、准确地掌握海岛海岸带资源分布,对其开发与利用、可持续发展具有重要意义。遥感技术由于其同步观测区域广、数据的客观性和可比性、现势性高、经济效益与社会效益较高的特性,已经广泛地应用于海岸带资源监测领域。研究如何利用遥感影像高精度地提取海岛海岸带地物信息成为遥感工作者关注的热点之一,也是本文的出发点。本文以青岛市红岛的三期SPOT5卫星影像为数据源,以SRTM3及黄岛港潮高数据为辅助数据,按照面向对象的基于知识的分类方法高精度地提取地物信息并进行动态变化分析研究。主要的研究工作如下:(1)在1:3000比例尺下矢量化获取三期遥感影像的道路信息,并将其作为专题信息参与多尺度分割,从而提高影像分割结果的实用性。(2)采用面向对象的多尺度分割方法完成遥感影像的最优分割。本文对多尺度分割中各参数(形状权重、紧致度、尺度参数)进行设置,得到了适宜于本次研究的参数,即在形状权重0.1、紧致度0.5,加入道路专题信息的前提下,建立三级分割网:道路专题信息层、尺度500分割层、尺度120分割层,其中:于道路专题信息层提取道路,于尺度500的分割层提取海水,于尺度120的分割层提取植被、湿地、建筑物等地物。(3)基于知识的分类完成遥感影像地物分类。通过一系列的实验研究,建立各类地物的知识:采用Density提取道路,GLCM Homogeneity、MNDWI提取海水,MNDVI、MNDWI提取湿地,NDVI提取植被,GLCM Dissimilarity、Mean SNIR提取建筑物。实验方法的地物分类精度高,分类结果与目视判读相一致,三期影像分类图的总体精度最高达97.80%,最低达到92.80%;卡帕系数最高达到0.970,最低达到0.900。(4)结合潮高数据从水边线推算海岸线。从面向对象的基于知识的分类图获取遥感影像的水边线,其边界连续,能够很好地划分湿地与海水;通过查阅黄岛港的潮高数据,得出砂质海岸的海岸线为与海水邻接的养殖池塘的边界,而基岩海岸的海岸线为其水边线。(5)采用多种动态变化分析方法,如总量分析、变化率分析、转移矩阵分析,对海岸线及三期遥感影像分类图进行动态变化分析,生成动态变化分析图表。结果表明海岸线总长度增加,其中:2007年比2003年增加727.9m,2011年比2007年增加1957.79m,但是,从2003年至2011年砂质海岸的岸线长度在不断减少而基岩海岸的岸线长度在不断增加;道路、在建区面积不断上升而湿地面积不断下降,湿地主要转换为建筑物及其他地物,是人为因素造成的;土地综合利用变化率高,其中2003至2007年土地综合利用变化率达到1.44%,而2007至2011年土地综合利用变化率达到1.95%。
[Abstract]:The coastal zone of the island is rich in resources and developed in economy. As a result of the development of the sea area, the urban expansion seriously causes the rapid change of land use in the coastal zone of the island. Therefore, it is of great significance to grasp the distribution of island coastal resources timely and accurately for its development and utilization and sustainable development. Remote sensing technology has been widely used in coastal resource monitoring field because of its wide synchronous observation area, objectivity and comparability of data, high present situation and high economic and social benefits. How to use the remote sensing image to extract the island coastal features information with high precision has become one of the hot topics of remote sensing researchers and is the starting point of this paper. In this paper, the third SPOT5 satellite image of Hongdao in Qingdao is taken as the data source, and the data of SRTM3 and the tidal height of Huangdao Port are taken as the auxiliary data. According to the object-oriented classification method based on knowledge, the ground object information is extracted with high precision and the dynamic change analysis is carried out. The main research work is as follows: (1) the road information of the third phase remote sensing image is vectorized at the scale of 1: 3000, and it is used as the thematic information to participate in multi-scale segmentation. In order to improve the practicability of image segmentation results. (2) the object oriented multi-scale segmentation method is used to complete the optimal segmentation of remote sensing images. In this paper, the parameters (shape weight, compactness, scale parameter) in multi-scale segmentation are set, and the parameters suitable for this study are obtained, that is, when the shape weight is 0.1, the compactness is 0.5, and the road thematic information is added, A three-level segmentation network was established: road thematic information layer, scale 500 segmentation layer, scale 120 segmentation layer, in which road was extracted from road thematic information layer, sea water was extracted from scale 500 partition layer, vegetation and wetland were extracted from scale 120 partition layer. (3) the classification of remote sensing images based on knowledge. Through a series of experimental studies, the knowledge of all kinds of ground objects was established: Density was used to extract GLCM homogenicity Density from sea water and wetland to extract vegetation. GLCM similarity and mean SNIR were used to extract buildings. The classification accuracy of the experimental method is high, and the classification results are consistent with visual interpretation. The overall accuracy of the third phase image classification map is 97.80, the lowest is 92.80, the kappa coefficient is 0.970 and the lowest is 0.900. (4) the shoreline is calculated from the water boundary with tidal height data. The waterlines of remote sensing images are obtained from object-oriented knowledge-based classification maps, whose boundaries are continuous and can well distinguish wetlands from sea water; by looking up the tidal height data of Huangdao Port, It is concluded that the shoreline of sandy coast is the boundary of culture pond adjacent to seawater, while the shoreline of bedrock coast is its water boundary. (5) A variety of dynamic change analysis methods, such as total quantity analysis, change rate analysis, transfer matrix analysis, are used. The dynamic change analysis of coastline and three stages remote sensing image classification map is carried out, and the dynamic change analysis chart is generated. The results show that the total length of shoreline has increased, including: 727.9 m more in 2007 than in 2003 and 1957.79 m in 2011 than in 2007. However, from 2003 to 2011, the length of shoreline of sandy coast is decreasing and the length of shoreline of bedrock coast is increasing. The area of the area under construction is rising and the area of the wetland is decreasing. The wetland is mainly converted into buildings and other features, which is caused by human factors, and the rate of change of comprehensive land use is high. From 2003 to 2007, the change rate of comprehensive land use reached 1.44%, while that of 2007 to 2011 reached 1.95%.
【学位授予单位】：中国石油大学(华东)
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：P715.7

【相似文献】