基于面向对象分类的右江区碳收支能力研究

发布时间：2018-06-28 17:44

本文选题：碳收支能力 + 多尺度分割　；参考：《广西师范学院》2016年硕士论文

【摘要】：全球气候变暖问题是人类面临的最艰巨的挑战之一,通过先进的面向对象分类方法对中、高分辨率光谱影像进行特征信息提取,利用影像信息估算区域碳收支能力,可以大大提高碳排放与碳汇能力的研究水平,对于控制区域气候变化具有推动作用。本文利用面向对象分类方法,以广西百色市右江区为研究区域,选取Landsat 8 OLI和Google Earth影像数据提取区域地物信息,并针对研究区地势复杂的特点,采用设置多种尺度参数的方法,选取最优尺度进行影像分割。同时,引入隶属度函数法、最邻近分类法和CART决策树分类器三种方法,基于影像光谱差异、几何形状、对象纹理等特征,逐层逐级地实施面向对象分类步骤,随后加入传统基于最大似然法的监督分类结果进行对比分析,针对性的实现分类结果精度评价分析用以检验分类结果的优势。通过总结分析前人的地物碳系数转换关系并结合高精度面向对象分类结果,构建基于土地覆被类型的碳收支能力估算模型,同时根据已有的基于CASA模型的碳收支能力估算方法加以精度校验,最终估算出右江区碳收支能力为-399.64万吨。本文还结合右江区行政区划、人口分布、DEM等相关数据对区域碳收支能力进行了专题性剖析。结果表明:(1)将RS和GIS技术应用与区域碳收支能力的研究具有很明显的优势。多尺度分割和面向对象分类方法能很大程度上消除光谱混淆引起的提取误差,提高中、高分辨率遥感影像的分类运行速度,并且解决了传统分类空间数据量大、分类结果具有“椒盐现象”等问题,使得不同分类方法可以对症下药,促使碳专题的分类精度有所提升。(2)归纳总结了国内外有关碳收支能力系数的研究成果,并将之应用于本研究构建碳收支估算模型当中。利用面向对象分类方法解译出的土地覆被数据和碳收支估算模型估算出右江区碳收支能力结果。结果表明林地、草地、耕地等植被覆盖程度高的地区担负了研究区的主要碳汇工作,而建设用地则因化石燃料的大批量消耗排放了大量的碳源,但右江区总体上碳汇能力比碳排放能力强,有利于区域生态系统的稳定发展,较为合理的缓解了区域气候变化的威胁。(3)结合右江区行政区划和DEM数据进行空间分析。总结出右江区碳收支能力总体特征为中心城区碳收支量大,四周乡镇碳收支量小。另一方面,研究分析出影响区域碳收支能力的主要因素有海拔、坡度坡向和人类活动等。具体表现为海拔高、坡度大、人类活动量少的地区碳汇量大、碳排放量小,而海拔低、坡度小、人类活动量大的地区碳排放量大、碳汇量小。综上所述,面向对象分类方法是研究小区域碳收支能力的有效途径,其分类效果好、速度快,精度高,在区域碳循环评估中具有更好的准确性和预见性,可以有效地促进碳收支平衡研究领域的发展。
[Abstract]:Global warming is one of the most difficult challenges facing mankind. Through advanced object-oriented classification methods, feature information is extracted from high-resolution spectral images, and the ability of estimating regional carbon budget is estimated by image information. It can greatly improve the research level of carbon emission and carbon sink ability, and promote the control of regional climate change. In this paper, the object oriented classification method is used to select Landsat 8 OLI and Google Earth image data to extract the regional feature information, taking Youjiang District, Baise City, Guangxi as the research area, and aiming at the complex features of terrain in the study area, the method of setting up various scale parameters is adopted. The optimal scale is selected for image segmentation. At the same time, three methods, membership function method, nearest neighbor classification method and cart decision tree classifier, are introduced. Based on image spectral difference, geometric shape, object texture and other features, the object oriented classification step by step is implemented step by step. Then the traditional supervised classification results based on the maximum likelihood method are compared and analyzed, and the accuracy evaluation analysis of the classification results is carried out to test the advantages of the classification results. Through summing up and analyzing the conversion relation of carbon coefficient of ground objects and combining with the result of high-precision object-oriented classification, a carbon budget estimation model based on land cover type is constructed. At the same time, according to the existing methods of estimating carbon budget and expenditure ability based on CASA model, the accuracy is checked, and finally, the carbon budget capacity of Youjiang region is estimated to be-3.9964 million tons. Based on the relevant data of administrative division and population distribution Dem in Youjiang District, the paper also analyzes the regional carbon budget and expenditure ability. The results show that: (1) the application of RS and GIS technology and the study of regional carbon budget have obvious advantages. Multi-scale segmentation and object-oriented classification method can eliminate the extraction error caused by spectral confusion to a great extent, improve the classification speed of middle and high resolution remote sensing images, and solve the problem of large amount of data in traditional classification space. The classification results have some problems, such as "salt and pepper phenomenon", which make different classification methods fit the case and promote the classification accuracy of carbon topics. (2) summarize the research results of carbon budget capacity coefficient at home and abroad. And it is applied in this study to build carbon budget estimation model. Based on the land cover data and carbon budget estimation model, the capacity of carbon budget in Youjiang region was estimated by using the object-oriented classification method. The results show that the areas with high vegetation coverage, such as woodland, grassland and cultivated land, are responsible for the main carbon sinks in the study area, while the construction land releases a large number of carbon sources due to the massive consumption of fossil fuels. But the ability of carbon sink in Youjiang area is stronger than that of carbon emission, which is beneficial to the stable development of regional ecosystem and reasonably alleviates the threat of regional climate change. (3) the spatial analysis of regional climate change is carried out in combination with administrative division and Dem data of Youjiang district. It is concluded that the overall characteristics of the carbon budget capacity in Youjiang District are that the carbon budget in the central urban area is large, and that in the surrounding towns and villages the carbon budget is small. On the other hand, the main factors influencing regional carbon budget are altitude, slope and human activities. The specific performance is high altitude, large slope, low human activity, large carbon sink, small carbon emission, but low elevation, small slope, large amount of carbon emission and small carbon sink in the area with large human activity. To sum up, the object-oriented classification method is an effective way to study the ability of carbon revenue and expenditure in small regions. It has good classification effect, fast speed and high precision, and has better accuracy and predictability in the assessment of regional carbon cycle. It can effectively promote the development of carbon balance research field.
【学位授予单位】：广西师范学院
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：X87

【相似文献】