基于多季相遥感信息的三江平原湿地信息提取

发布时间：2018-05-04 08:48

本文选题：面向对象的方法 + 多季相Landsat　；参考：《中国科学院研究生院(东北地理与农业生态研究所)》2013年硕士论文

【摘要】：本文以中国东北地区三江平原为研究对象，利用研究区2000年、2006年和2012年多季相Landsat TM/ETM+影像，建立了湿地分类系统，结合野外调查数据，应用多尺度分割算法，根据影像的丰富信息和物候、时相等特征，采用面向对象的分类方法，进行三江平原的湿地分类。且利用遥感解译数据，逐步分析了研究区2000-2012年湿地的覆被空间特征，并分析了湿地类型的变化特征。结果表明： 1.2000年、2006年和2012年湿地提取的生产者精度和用户精度都高于85%，总体精度也都高于85%。面向对象的分类方法能够有效利用遥感影像提供的丰富信息，产生较高的分类精度。由于湿地分布破碎复杂，且中分辨率影像地物边界模糊，影像分割时，尺度太大难以准确反映湿地分布状况和边界，分割尺度过小则不利于对象的提取和空间信息的利用。对于中分辨率遥感影像不同分割尺度的组合使用有利于湿地信息的提取。 2.多季相影像数据的使用，改善了季相变化湿地的分类效果和精度，要获得较高的湿地分类精，降雨量多寡月份的选取和物候规律的把握是关键。针对特定的年份，最好选取当年降水最大和最小的月份，且结合能够明显区分植被季相信息月份的遥感影像。利用面向对象使用中分辨率多季遥感影像，在大区域范围内细致且高精度的提取湿地信息是一个成本较低且行之有效的办法。 3.以季节变化为时间尺度，同年各季相遥感影像为解译目标，发现降水较多季节和降水较少季节之间影像监测的湿地转化最为剧烈，主要为湿地类型到非湿地类型之间的转化，湿地类型内部间的转化和非湿地类型到湿地类型的转化所占面积比例均很小。季节性的湿地转化主要发生在草本沼泽和草地之间。降水、气温及植被长势等变化因素对湿地面积季相变化及转化起到很大作用。 4.三江平原湿地主要分布在平原、低洼地区。主要人工湿地为水田，其余人工湿地面积比例较少；主要天然湿地类型为草本沼泽，其次为河流，其他湿地类型所占面积比例较小。湿地分布主要在保护区、河岸以及边远国界周围，而且分布较为集中。研究区2000，2006和2012年各阶段湿地覆被变化明显，，主要发生在湿地和非湿地之间，湿地内部转化次之。湿地总体面积稳定，近十年三江平原草本沼泽农田化进程趋缓。
[Abstract]:This paper takes Sanjiang Plain in Northeast China as the research object, using the multi-season Landsat TM/ETM images in 2000, 2006 and 2012, establishes the wetland classification system, combines the field survey data, and applies the multi-scale segmentation algorithm. According to abundant information, phenology and time-equal features of images, the wetland classification of Sanjiang Plain is carried out by using object-oriented classification method. The spatial characteristics of wetland cover and the variation characteristics of wetland types in the study area from 2000 to 2012 were analyzed by means of remote sensing interpretation data. The results show that: 1. In 2000, 2006 and 2012, the producer and user accuracy of wetland extraction was higher than 850.The overall precision was also higher than that of 85kyam. The object-oriented classification method can effectively utilize the abundant information provided by remote sensing images and produce higher classification accuracy. Because the distribution of wetland is complicated and the boundary of middle resolution image is blurred, it is difficult to accurately reflect the distribution and boundary of wetland when image segmentation is too large. The small scale of segmentation is not conducive to the extraction of objects and the use of spatial information. The combination of different segmentation scales for middle resolution remote sensing images is beneficial to the extraction of wetland information. 2. The use of multi-season image data improves the classification effect and precision of seasonal wetland. To obtain higher wetland classification precision, the selection of rainfall month and the grasp of phenological regularity are the key. For a specific year, it is better to select the month with the largest and the smallest precipitation in the year, and combine the remote sensing image which can distinguish the vegetation seasonal information month obviously. It is a low cost and effective method to extract wetland information carefully and accurately in large area by using object oriented multi-season remote sensing image with medium resolution. 3. With seasonal variation as time scale and remote sensing images of different seasons as the interpretation target in the same year, it was found that the transformation of wetlands monitored by image between more and less precipitation seasons was the most intense, mainly the transformation from wetland type to non-wetland type. The area ratio of the transformation between the wetland types and the non-wetland types to the wetland types is very small. Seasonal wetland conversion occurs mainly between herbaceous swamps and grasslands. Precipitation, air temperature and vegetation growth play an important role in the seasonal change and transformation of wetland area. 4. Sanjiang plain wetland mainly distributes in plain, low-lying area. The main artificial wetland is paddy field, the other constructed wetland area proportion is less, the main natural wetland type is herbaceous swamp, the next is river, the other wetland type accounts for a small proportion of the area. Wetlands are mainly distributed in protected areas, riverbanks and around remote borders, and the distribution is relatively concentrated. The changes of wetland cover in 2006 and 2012 were obvious, mainly occurred between wetland and non-wetland, followed by internal transformation of wetland. The total area of wetland is stable, and the farmland process of herbaceous swamp in Sanjiang Plain has slowed down in recent ten years.
【学位授予单位】：中国科学院研究生院(东北地理与农业生态研究所)
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：P237

【参考文献】