【作者】 李开丽；

【导师】 倪绍祥； 蒋建军；

【作者基本信息】 南京师范大学 ， 地图学与地理信息系统， 2005， 硕士

【摘要】 东亚飞蝗是引发我国蝗灾的最常见蝗虫。我国史籍中记载的800多次蝗灾,主要是指东亚飞蝗所引起的蝗灾。近年来,由于气候变化和人类活动等原因,这类蝗灾在我国有不断加重之趋势。 及时、有效地防治蝗灾的关键,是对蝗灾作出监测;而现代遥感技术因其具有大面积快速、多时相等优点,因此在蝗灾的监测中可发挥特别的作用。 蝗虫的发生与成灾与其存在的生境有十分密切的关系。因此,对蝗虫生境进行监测便成为遥感技术用于蝗灾监测的重要内容;而基于遥感影像的生境的计算机分类是蝗灾遥感监测的一项重要基础性工作。 本文以渤海湾沿岸的河北省黄骅地区为研究区,进行东亚飞蝗生境的遥感分类研究。首先,对东亚飞蝗生境的含义进行了探讨。其次,结合黄骅地区的实际情况和所采用卫星影像特征,确定了东亚飞蝗生境分类的原则,并以此原则建立了黄骅地区东亚飞蝗的生境分类系统。然后,分别利用包括2003年8月14日和2004年5月28日的Landsat-5 TM图像,以及2003年10月16日的ASTER图像,采用最大似然法和基于知识的分层分类法,对黄骅地区的东亚飞蝗生境进行了不同组合的分类研究,并经过分析和比较,找出了最适合研究区东亚飞蝗生境分类的组合方案。 通过本文研究,可得到以下结论: 1) 正确地使用基于知识的分层分类法,能在很大程度上提高东亚飞蝗生境的分类精度。最大似然法也是一种很好的生境分类方法,但在分类中应注意训练样本的选取和纯化。 2) 辅以空间信息的遥感影像生境分类,可使分类精度有所提高。 3) 充分利用多时相遥感信息,可在很大程度上提高生境分类的精度。研究表明:分别将2003年8月14日和2004年5月28日的TM影像的5、4、3波段进行重组,其最大似然法分类的总精度可达到88.8095%,Kappa系数达0.8462;而增加了空间信息后的分类总精度可达89.3188%,Kappa系数达0.8534。 4) ASTER影像的利用使生境分类精度有进一步的提高。结果表明,对2003年8月14日TM影像和2003年10月16日的ASTER影像的融合图像进行基于知识的分层分类,分类总精度可达90.9739%,Kappa系数达0.8738。此分类精度可基本满足黄骅地区东亚飞蝗生境遥感分类与制图的目的。更多还原

【Abstract】 Oriental migratory locust (Locusta migratoria manilensis Meyen) is the kind of locust which has very often brought the harmful effects to many regions in China. More than 800 locust plagues have been recorded in Chinese historical documents, in which most of the plagues were caused by oriental migratory locust. In recent years, a tendency of more and more serious locust plague has emerged due to the climatic change and human activities.The key of timely and efficiently control of locust plague is conducting the monitoring of the plague. Remote sensing as the modern technology and with the advantages such as large-area and multi-temporal monitoring has a particular role in locust plague monitoring.There are close relationships between locust occurrence and outbreak and their habitas. Therefore, the monitoring of locust habitats has became a very important task in the application of remote sensing in locust plague monitoring, and habitat classification based on remotely sensed imagery has became an important foundamental task.In this paper, Huanghua city near Bohai Bay in Hebei province was taken as the study area and the locust habitat classification based on the images was conducted. First, the meanings of habitat of oriental migratory locust were explored. Second, the priciples used for the classification were determined based on the practical conditions of the study area and the used remotely sensed images, and a habitat classification system was constructed according to the classification principles. Third, the diffent combinations of the habitat classification were tested, including using the images of Landsat-5 received on August 14, 2003 and May 28 of 2004, and ASTER received at October 16 of 2003, and using the different classification methods including the maximum likehood classifier and knowleage-based layered classifier. Through the analysis and comparison of the combinations the mostly suitable one to the habitat classification in the study area was determined.Through this study the follow conclusions are drawn:1) Correctly using the knowleage-based layered classifier can much improve the classification accuracy of locust habitats. The maximum likelihood classifier is a rather good classification method and, however, much attention should be paid to selecting and purification of the training samples.2) With the aid of the spatial texture information on image, the accuracy of the habitat classification will be more or less improved.3) The accuracy of habitat classification will be particularly improved if multi-temporal remotely sensed image data are applied. The study indicated that: the overall accuracy and Kappacoefficient of the maximum likelihood are 88.8095% and 0.8462 respectively if combining of TM 5,4,3 bands of August 14, 2003 and May 28, 2004; and the overall accuracy and Kappa coefficient of the maximum likelihood will arise to 89.3188% and 0.8534 respectively if adding the spatial texture information on the images.4) Using ASTER image data can further improve the accuracy of the habitat classification. The study has shown that the overall accuracy and Kappa coefficient of the classification is 90.9739% and 0.8738 respectively if using the fused images of TM data received on August 14, 2003 and the ASTER data on October 16, 2003, and using the knowledge-based layered classification. This accuracy can basically meet the demands of the classification and mapping of the habitats for oriental migratory locust in the study area.更多还原