本文为奥地利萨尔茨堡大学（作者：Bacc.ing.tech.inf. Zlatko Horvat）的硕士论文，共91页。

及时、准确地探测地表土地利用-土地覆盖变化信息，对于了解人类与自然现象之间的关系和相互作用、更好地进行决策管理具有极其重要的意义。本研究利用Landsat卫星图像评价了1978年、1992年至2007年间克罗地亚MEĐIMURJE县LULC的变化。利用MSS、TM和ETM+三幅Landsat卫星图像对LULC变化的空间动态进行了量化。制作了1978年、1992年和2007年三个系列的土地利用变化趋势图。采用基于“爬山”聚类算法的无监督分类、基于最大似然分类法（MLC）的监督分类和从实地经验引入的人类知识，在MEĐIMURJE县的所有城市地区进行了1:25000地形图和高空摄影的仔细手工划定。之后，从图像数据中剔除城市区域，并用无监督分类法将其与其他数据分开分类。根据安德森分类系统，土地利用和土地覆盖分为：水体、林地、荒地、农业用地和城市建设用地。误差矩阵被用来评估分类的准确性。总体精度、用户和生产商的精度以及Kappa统计量均由误差矩阵导出。1978年、1992年和2007年的总准确率分别为87.67%、88.96%和90.84%。Kappa统计分别为80%、84%和83%。用户和生产商对各个类别的准确度相对较高，从74%到94%不等，这表明由图像生成的专题地图与参考数据之间有很好的一致性。分析结果表明，从1978到2007年，城市面积增加了约53%，而农业用地减少了12%。水、林、荒地分别增加14%、5%和49%。导出的MEĐIMURJE县土地利用/土地覆盖变化图提供了研究期间土地变化的良好信息。  

Timely and precise information about LandUse Land Cover (LULC) change detection of the Earth’s surface is extremelyimportant for understanding relationships and interactions between human andnatural phenomena for better management of decision making. This studyevaluates LULC changes in Međimurje County, Croatia, between 1978, 1992 and2007 using Landsat satellite images. Spatial dynamics of LULC changes werequantified using three Landsat satellite images: MSS, TM and ETM+ . Threeseries of the LULC maps of 1978, 1992 and 2007 were produced. A postclassification technique was used based on a hybrid classification approachcomprised of unsupervised classification based on “hill climbing” clusteralgorithm, supervised classification based on the Maximum Likelihood Clasifier(MLC) and introduction of human-knowledge from field experience. Careful manualdelineation on the topographic maps 1:25,000 and high aerial photographs weredone around all urban areas of Međimurje County. The urban areas were clippedout of the image data later and classified separately from other data with anunsupervised classification. Based on the Anderson classification system, theland-use and land- covers are classified as: water bodies, forest land, barrenland, agricultural land and urban built up land. Error matrices were used toassess classification accuracy. The overall accuracies, user’s and producer’saccuracies, and the Kappa statistics were derived from the error matrices. Theoverall accuracies for 1978, 1992 and 2007 were, respectively, 87.67%, 88.96%and 90.84%. Kappa statistics were 80%, 84% and 83%. User's and producer's accuraciesof the individual classes were relatively high, ranging from 74% to 94% whichindicates a good agreement between the thematic maps generated from images andthe reference data. The results of the analysis showed that from 1978 to 2007the urban area increased approximately 53%, while the agricultural landdecreased 12%. Water, forest and barren land also increased 14%, 5% and 49%respectively. The derived LULC changes maps of Međimurje County provide goodgeneral information about land changes during the study period.

1. 引言

2. 文献回顾

3. 研究区域、数据描述和使用的软件

4. 研究方法

5. 结果与讨论

6. 总结、结论与建议

更多精彩文章请关注公众号：