基于马尔科夫随机场与模糊C均值的遥感影像分类

模糊C均值聚类是一种经典的非监督聚类模型,成功地应用于遥感影像分类。但是该方法对初始值敏感,容易陷入局部最优解;同时聚类时仅考虑光谱信息,忽略了空间信息。本文提出了一种新的基于马尔科夫随机场的模糊C均值聚类方法,该方法把马尔科夫随机场和模糊C均值结合在一起。初始值依据第一主成分的密度函数确定,这样克服了对初始值的依赖性,又在聚类的时候考虑了空间信息。通过实例数据验证,所提出的方法分类精度优于传统的模糊C均值模型。     

融合多特征的遥感影像变化检测

常规多时相遥感影像变化检测主要基于光谱信息,没有充分利用纹理、几何、形状等多种特征信息,不足以体现检测目标的完整性和准确性。本文针对不同特征在变化检测中应用的优势,在提取影像多种特征的基础上,构建了1维和多维两种基于信息融合策略的变化检测方法,即利用1维特征空间加权距离相似度运算、多维特征空间的模糊集融合和支持向量机融合策略进行变化检测。利用多时相QuickBird高分辨率遥感影像进行城市土地覆盖变化检测试验,结果表明,本文方法可以有效集成不同特征的优势与表征变化信息的能力,提高变化检测过程的稳定性和适用性,同时能够更好地保持变化地物的结构和形状,突出主要变化目标。     

一种自动搜索水准环及计算闭合差的方法研究

本文采用深度优先搜索算法(Depth-First-Search)实现自动搜索水准闭合环,并结合ArcGIS的图形拓扑关系技术来获得最小水准环的信息;将此项技术应用于水准测量闭合环的闭合差计算,最终实现了自动计算水准闭合环的闭合差及精度统计。     

模糊综合评价法在新疆农业主导功能定位中的应用

 新疆区域农业主导功能具有空间和组合上的复合性特征,探讨建立一种适用于区域农业主导功能为复合功能,并对复合型农业主导功能的主要功能和辅助功能进行明确定位的综合评价方法,对保证新疆农业功能区划的科学性、合理性和可操作性具有十分重要的意义。在研究构建新疆农业功能综合评价指标体系的基础上,应用模糊综合评价方法,构建综合评价模型,对全疆88 个涉农县（市、区）的农业功能进行综合评价。在综合测评结果基础上,对各县（市、区）的农业功能重要性进行纵向比较和横向比较,以确定各区划单元四大农业功能的主辅地位以及优势主导功能。根据模糊综合评价的测评排序结果,结合问卷调查和专家咨询结果,提出了新疆区域农业主导功能定位及布局方案。应用模糊综合评价方法对新疆农业功能进行综合测评和主导功能定位,同时结合定性方法对评价结果进行比对和修正,得出的研究方案符合新疆实际,表明该方法是对复合型农业主导功能区进行农业功能定位及区划的有效方法之一,同时为新疆农业主导功能区制定不同的发展对策和战略措施,优化新疆农业产业结构和布局提供了科学依据。     

基于局部空间统计的产业集群空间分析——以北京市制造业集群为例

将集群识别与空间分析相结合,通过全国集群模板识别产业集群,进而利用局部空间统计方法,探测产业集群的空间布局特征。利用2008年北京市和全国经济普查数据,从产业联系的角度识别了北京市制造业集群,在此基础上进一步测度了北京市制造业集群的空间分布现状与布局特征。研究发现,北京市制造业集群主要分布在近郊区并向远郊区延伸,资源条件、政策环境、交通区位、历史因素等多种区位因子均对集群的区位选择产生影响。从总体分布来看,北京市制造业集群的空间布局存在分布比较分散、功能重叠或与区县功能定位不符等问题。未来在政策制定时,应进一步强化北京市制造业集群的空间集聚,发挥区域产业政策的引导作用,并不断依托产业园区优化集群发展环境,引导符合区县功能定位的制造业集群发展。     

The DEEP2 Galaxy Redshift Survey: the red sequence AGN fraction and its environment and redshift dependence

We measure the dependence of the active galactic nuclei (AGN) fraction on local environment at   z ∼ 1  , using spectroscopic data taken from the DEEP2 Galaxy Redshift Survey, and Chandra X-ray data from the All-Wavelength Extended Groth Strip International Survey (AEGIS). To provide a clean sample of AGN, we restrict our analysis to the red sequence population; this also reduces additional colour–environment correlations. We find evidence that high-redshift LINERs in DEEP2 tend to favour higher density environments relative to the red population from which they are drawn. In contrast, Seyferts and X-ray selected AGN at   z ∼ 1  show little (or no) environmental dependencies within the same underlying population. We compare these results with a sample of local AGN drawn from the Sloan Digital Sky Survey (SDSS). Contrary to the high-redshift behaviour, we find that both LINERs and Seyferts in the SDSS show a slowly declining red sequence AGN fraction towards high-density environments. Interestingly, at   z ∼ 1  red sequence Seyferts and LINERs are approximately equally abundant. By   z ∼ 0  , however, the red Seyfert population has declined relative to the LINER population by over a factor of ∼4.5. We speculate on possible interpretations of our results.     

Compact groups in theory and practice – II. Comparing the observed and predicted nature of galaxies in compact groups

We examine the properties of galaxies in compact groups (CGs) identified in a mock galaxy catalogue based upon the Millennium Run simulation. The overall properties of groups identified in projection are in general agreement with the best available observational constraints. However, only ∼30 per cent of these simulated groups are found to be truly compact in three dimensions, suggesting that interlopers strongly affect our observed understanding of the properties of galaxies in CGs. These simulations predict that genuine CG galaxies are an extremely homogeneous population, confined nearly exclusively to the red sequence: they are best described as 'red and dead' ellipticals. When interlopers are included, the population becomes much more heterogeneous, due to bluer, star-forming, gas-rich, late-type galaxies incorrectly identified as CG members. These models suggest that selection of members by redshift, such that the line-of-sight velocity dispersion of the group is less than 1000 km s⁻¹, significantly reduces contamination to the 30 per cent level. Selection of members by galaxy colour, a technique used frequently for galaxy clusters, is also predicted to dramatically reduce contamination rates for CG studies.     

Strong lensing of submillimetre galaxies: a tracer of foreground structure?

The steep source counts and negative K -corrections of bright submillimetre galaxies (SMGs) suggest that a significant fraction of those observed at high flux densities may be gravitationally lensed, and that the lensing objects may often lie at redshifts above 1, where clusters of galaxies are difficult to detect through other means. In this case, follow-up of bright SMGs may be used to identify dense structures along the line-of-sight. Here, we investigate the probability for SMGs to experience strong lensing, using the latest N -body simulations and observed source flux and redshift distributions. We find that almost all high-redshift sources with a flux density above 100 mJy will be lensed if they are not relatively local galaxies. We also give estimates of the fraction of sources experiencing strong lensing as a function of observed flux density. This has implications for planning follow-up observations for bright SMGs discovered in future surveys with the Submillimetre Common-User Bolometer Array 2 and other instruments. The largest uncertainty in these calculations is the maximum allowed lensing amplification, which is dominated by the presently unknown spatial extent of SMGs.