Deforestation in Central Africa: Estimates at regional, national and landscape levels by advanced processing of systematically-distributed Landsat extracts

Accurate land cover change estimates are among the headline indicators set by the Convention on Biological Diversity to evaluate the progress toward its 2010 target concerning habitat conservation. Tropical deforestation is of prime interest since it threatens the terrestrial biomes hosting the highest levels of biodiversity. Local forest change dynamics, detected over very large extents, are necessary to derive regional and national figures for multilateral environmental agreements and sustainable forest management. Current deforestation estimates in Central Africa are derived either from coarse to medium resolution imagery or from wall-to-wall coverage of limited areas. Whereas the first approach cannot detect small forest changes widely spread across a landscape, operational costs limit the mapping extent in the second approach. This research developed and implemented a new cost-effective approach to derive area estimates of land cover change by combining a systematic regional sampling scheme based on high spatial resolution imagery with object-based unsupervised classification techniques. A multi-date segmentation is obtained by grouping pixels with similar land cover change trajectories which are then classified by unsupervised procedures. The interactive part of the processing chain is therefore limited to land cover class labelling of object clusters. The combination of automated image processing and interactive labelling renders this method cost-efficient. The approach was operationally applied to the entire Congo River basin to accurately estimate deforestation at regional, national and landscape levels. The survey was composed of 10 × 10 km sampling sites systematically-distributed every 0.5° over the whole forest domain of Central Africa, corresponding to a sampling rate of 3.3%. For each of the 571 sites, subsets were extracted from both Landsat TM and ETM+ imagery acquired in 1990 and 2000 respectively. Approximately 60% of the 390 cloud-free samples do not show any forest cover change. For the other 165 sites, the results are depicted by a change matrix for every sample site describing four land cover change processes: deforestation, reforestation, forest degradation and forest recovery. This unique exercise estimates the deforestation rate at 0.21% per year, while the forest degradation rate is close to 0.15% per year. However, these figures are less reliable for the coastal region where there is a lack of cloud-free imagery. The results also show that the Landscapes designated after 2000 as high priority conservation zones by the Congo Basin Forest Partnership had undergone significantly less deforestation and forest degradation between 1990 and 2000 than the rest of the Central African forest.     

基于半监督多示例学习的对象图像检索

针对基于对象的图像检索问题,提出一种新的半监督多示例学习(MIL)算法.该算法将图像当作包,分割区域的视觉特征当作包中的示例,按"点密度"最大原则,提取"视觉语义"构造投影空间;然后利用定义的非线性函数将包映射成投影空间中的一个点,以获得图像的"投影特征",并采用粗糙集(RS)方法对其进行属性约简;最后利用直推式支持向量机(TSVM)进行半监督的学习,得到分类器.实验结果表明,该方法有效且性能优于其他方法.     

The use of high-resolution remote sensing for plague surveillance in Kazakhstan

Bubonic plague, caused by the bacteria Yersinia pestis, persists as a public health problem in many parts of the world, including central Kazakhstan. Bubonic plague occurs most often in humans through a flea bite, when a questing flea fails to find a rodent host. For many of the plague foci in Kazakhstan the great gerbil is the major host of plague, a social rodent well-adapted to desert environments. Intensive monitoring and prevention of plague in gerbils started in 1947, reducing the number of human cases and mortalities drastically. However, the monitoring is labour-intensive and hence expensive and is now under threat due to financial constraints. Previous research showed that the occupancy rate of the burrow systems of the great gerbil is a strong indicator for the probability of a plague outbreak. The burrow systems are around 30 m in diameter with a bare surface. This paper aims to demonstrate the automatic classification of burrow systems in satellite images using object-oriented analysis. We performed field campaigns in September 2007 and May and September 2008 and acquired corresponding QuickBird images of the first two periods. User's and producer's accuracy values of the classification reached 60 and 86%, respectively, providing proof of concept that automatic mapping of burrow systems using high-resolution satellite images is possible. Such maps, by better defining great gerbil foci, locating new or expanding foci and measuring the density of great gerbil burrow systems could play a major role in a renewed monitoring system by better directing surveillance and control efforts. Furthermore, if similar analyses can separate occupied burrow systems from empty ones, then very-high-resolution images stand to play a crucial role in plague surveillance throughout central Asia.     

Cfslight:一个新型的轻量级对象存储集群文件系统

对象存储技术是国际上提出的集群文件系统的新技术，已成为国际上的研究热点。文章基于Linux操作系统实现了一个新型的轻量级对象存储集群文件系统Cfslight，它的优点是配置安装简单，使用方便．大大提高了集群系统的易用性、易管理性。此外，由于采用了对象存储技术的核心思想和多线程技术以及对象存储体在操作系统内核实现，因而Cfslight还具有较好的性能。     

Segmentation of High Spatial Resolution Remotely Sensed Data Using Watershed with Self-adaptive Parameterization

Segmentation of high spatial resolution remotely sensed image is the important foundation of Object\|Based Image Analysis(OBIA), most of the image segmentation algorithms involve the problem of parameter setting. Self\|adaptive Parameterization is one of the key factors that affect the efficiency and effectiveness of remote sensing image segmentation. Considering that traditional watershed segmentation algorithm is susceptible to noise and the segmentation scale parameter is difficult to be self\|adaptively chosen,this paper propose a scale self\|adaptive method in watershed segmentation. After median filtering in primary image, this paper uses spatial statistical method to realize the self\|adaptive setting of watershed segmentation parameters, and then segments the high spatial resolution remote sensing image. This study uses IKONOS and Quickbird multispectral images as experimental data to testify the validity of the method proposed by this paper. The homogeneity within the segmentation parcels and the heterogeneity between thesegmentation parcels are used to build up a synthetic evaluation model to quantitatively evaluate the segmentation results by the proposed method by comparing with different parameter sequences segmentation results. The comparison result show that the proposed method perform well in high spatial resolution image segmentation. As result, the method proposed in this paper not only improves the accuracy of image segmentation to a certain extent, but also raises the automation of the segmentation parameter selection, which provides a new way for image segmentation and the research of parameterization in the future.     

遥感图像多尺度分割算法与矢量化算法的集成

在遥感图像处理和分析软件中,图像分割/分类和矢量化是前后独立的过程:先分割/分类,再矢量化整幅图像。由于矢量化后得到的矢量文件未写入图像对象(区域、图斑)的特征信息,只能用于显示而不能用于后续操作。此外,处理复杂图像时还存在矢量文件多边形数目与分割/分类后图像区域数目不一致的问题。为此,将多尺度分割算法与矢量化算法进行一体化集成。对遥感图像进行多尺度分割得到图像对象链表,逐个对图像对象做矢量化处理,同时把特征统计信息写入多边形属性中。集成后不仅可保证矢量多边形数目与图像对象数目完全一致,而且由于特征统计信息已作为多边形区域的属性保存在多边形中,后续的多尺度分割、区域合并、空间关系操作等均可基于矢量多边形进行。     

2-D mesh-based video object segmentation and tracking with occlusion resolution

This paper integrates fully automatic video object segmentation and tracking including detection and assignment of uncovered regions in a 2-D mesh-based framework. Particular contributions of this work are (i) a novel video object segmentation method that is posed as a constrained maximum contrast path search problem along the edges of a 2-D triangular mesh, and (ii) a 2-D mesh-based uncovered region detection method along the object boundary as well as within the object. At the first frame, an optimal number of feature points are selected as nodes of a 2-D content-based mesh. These points are classified as moving (foreground) and stationary nodes based on multi-frame node motion analysis, yielding a coarse estimate of the foreground object boundary. Color differences across triangles near the coarse boundary are employed for a maximum contrast path search along the edges of the 2-D mesh to refine the boundary of the video object. Next, we propagate the refined boundary to the subsequent frame by using motion vectors of the node points to form the coarse boundary at the next frame. We detect occluded regions by using motion-compensated frame differences and range filtered edge maps. The boundaries of detected uncovered regions are then refined by using the search procedure. These regions are either appended to the foreground object or tracked as new objects. The segmentation procedure is re-initialized when unreliable motion vectors exceed a certain number. The proposed scheme is demonstrated on several video sequences.     

一种用于MPEG-4形状编码的快速运动估计算法

该文提出了一种用于MPEG-4形状编码的快速运动估计算法。该算法利用了形状编码及形状信息的固有特性,即基于上下文的运动估计特性、相邻二值alpha块的运动矢量相关性以及形状信息的二值特性。模拟结果表明,该算法具有运算量少、处理速度快的特点,适用于MPEG-4形状编码的实时软件实现。     

基于高分辨率遥感影像的土地覆盖信息提取

高空间分辨率遥感影像使得土地覆盖和土地利用信息的提取成为可能。以高分辨率遥感影像数据IKONOS为主要数据源,以多尺度分割与基于模糊逻辑分类的面向对象影像分析方法为主要技术,自动提取株洲市城乡结合部的土地覆盖和土地利用信息。达到了提取郊区丘陵地带林地信息和城市建筑、道路等土地覆盖信息的目的,而且精度高,速度快。结果表明利用该方法对复杂的城乡结合部信息获取是可行的。     

Estimation of point light source parameters for object-based coding

In this paper, the source model of moving rigid 3D objects of an object-based analysis-synthesis coder (OBASC) is extended from diffuse to non-diffuse illumination introducing the explicit illumination model of a distant point light source and ambient diffuse light. For each image of a real image sequence containing moving objects, first, shape and 3D motion parameters describing the objects are estimated assuming an ellipsoid-like smooth shape. Then, the illumination parameters are estimated by a fast iterative maximum-likelihood Gauβ-Newton estimation method. Typically, the illumination parameters converge after very few images close to the true ones. The accurateness depends on the amount of object rotation and the correctness of the shape assumptions. For a real image sequence showing a textured ball covering 20% of image area, rotating about 10 ° per frame, and illuminated by spot and ambient light, the extension of the source model reduces the model failures from 9.9% of the image area to 6.7%. In the area of model failures, the image synthesized from the source model parameters differ significantly from the real image. In this early experiment, source model parameters are coded losslessly. Since model failures are expensive by means of bit-rate, a significant reduction of bit-rate can be expected.