Unconstrained monocular 3D human pose estimation by action detection and cross-modality regression forest

This work addresses the challenging problem of unconstrained 3D human pose estimation (HPE) from a novel perspective. Existing approaches struggle to operate in realistic applications, mainly due to their scene-dependent priors, such as background segmentation and multi-camera network, which restrict their use in unconstrained environments. We therfore present a framework which applies action detection and 2D pose estimation techniques to infer 3D poses in an unconstrained video. Action detection offers spatiotemporal priors to 3D human pose estimation by both recognising and localising actions in space-time. Instead of holistic features, e.g. silhouettes, we leverage the flexibility of deformable part model to detect 2D body parts as a feature to estimate 3D poses. A new unconstrained pose dataset has been collected to justify the feasibility of our method, which demonstrated promising results, significantly outperforming the relevant state-of-the-arts. © 2013 IEEE.

EXPERIMENTS ON AQUATIC ECOSYSTEM REGIME SHIFT IN ENCLOSURES NEAR LAKE DIANCHI, CHINA

To discover how a lake converts from a turbid state to clean state, and what drives this process, we constructed controlled enclosure ecosystems and used the ecological remediation method to force ecosystems to convert from the turbid state to the clean state. Our results show that the driving forces include temperature., macrophyte, silver carp and mussel, which form a combined force to drive the controlled ecosystem to switch. There is a threshold existing in treated enclosure ecosystem during the conversion from turbid to clean state. When TP <0.09 mg.L-1, Chl-a <0.036 mg.L-1, transparency >62 cm, TN <2.15 mg.L-1, CODMn <13.7 mg.L-1, tubidity <10, and the number of algal cells <10(6) cells.L-1, the treated ecosystem changes sharply from turbid to clean state. The conversion process can be divided into three phases: turbid state, clean-turbid transitional state as well as clean state, and described with the power function Y = a*X-b (where Y is water parameter, X is time, a and b are constants), which indicates that the shift in the enclosure ecosystem from turbid to clean state is discontinuous.

The Random Forest Kernel and other kernels for big data from random partitions

We present Random Partition Kernels, a new class of kernels derived by demonstrating a natural connection between random partitions of objects and kernels between those objects. We show how the construction can be used to create kernels from methods that would not normally be viewed as random partitions, such as Random Forest. To demonstrate the potential of this method, we propose two new kernels, the Random Forest Kernel and the Fast Cluster Kernel, and show that these kernels consistently outperform standard kernels on problems involving real-world datasets. Finally, we show how the form of these kernels lend themselves to a natural approximation that is appropriate for certain big data problems, allowing $O(N)$ inference in methods such as Gaussian Processes, Support Vector Machines and Kernel PCA.

A simple closed aquatic ecosystem (CAES) for space

A closed aquatic ecosystem (CAES) was developed to stud), the effects of microgravity on the function of closed ecosystems aboard the Chinese retrieved satellite and on the spacecraft SHENZHOU-II. These systems housed a small freshwater snail (Bulinus australianus) and an autotrophic green algae (Chlorella pyrenoidosa). The results of the test on the satellite were that the concentration of algae changed little, but that the snails died during the experiments. We then sought to optimize the function of the control system, the cultural conditions and the data acquisition system and carried out an experiment on the spacecraft SHENZHOU-II. Using various sensors to monitor the CAES, real-time data regarding the operation of the CAES in microgravity was acquired. In addition, all on-board Ig centrifuge was included to identify gravity-related factors. It was found that microgravity is the major factor affecting the operation of the CAES in space. The change in biomass of the primary producer during each day in microgravity was larger than that of the control groups. The mean biomass concentration per day in the microgravity group decreased, but that of the control groups increased for several days and then leveled off. Space effects on the biomass of a primary producer may be a result of microgravity effects leading to increasing metabolic rates of the consumer combined with decreases in photosynthesis. (c) 2007 COSPAR. Published by Elsevier Ltd. All rights reserved.

长白山阔叶红松林生态系统碳密度对采伐干扰的响应研究

本研究以长白山地区原始和次生的阔叶红松林为对象，在2007和2008年共建立了8种类型10块1 ha样地。通过野外调查和取样分析，得到各样地森林生态系统的植物、枯死物和土壤碳密度值，并结合采伐样地经营历史情况（采伐时间和强度），得到长白山地区原始阔叶红松林生态系统碳密度参考值和次生阔叶红松林生态系统碳密度对采伐强度和植被恢复时间的响应特征。在此基础上，通过建立阔叶红松林乔木碳密度与生态系统碳密度的回归关系，并结合露水河林业局3个时期（1987、1995和2003年）的小班数据和原始林碳密度参考值，估算露水河林业局林业用地3个时期和潜在的碳储量。主要研究结果如下： （1）原始阔叶红松林生态系统碳密度参考值：植被、枯死物、土壤和生态系统碳密度值分别为149.18±54.57、20.93±14.33、156.39±14.99、326.50±34.52 t•ha-1。其中下木层、乔木层碳密度分别是1.55±0.74，147.63±54.39 t•ha-1；粗木质残体和枯枝落叶碳密度分别是15.64±13.66、5.29±1.72 t•ha-1；0-50 cm的各层土壤碳密度分别为62.14±6.31、46.17±10.25、27.82±6.20、12.57±4.67、7.69±2.20 t•ha-1。 （2）原始阔叶红松林生态系统碳密度对采伐干扰的响应特征为：采伐干扰均会减少生态系统碳密度；其中植被碳库对采伐干扰最为敏感且碳密度值均减少，采伐强度直接决定植被碳密度的减少程度和恢复时间；枯死物碳库对采伐干扰最不敏感且碳密度值是先增加后减少，采伐强度和植被恢复时建群树种决定枯死物碳密度的变化程度；土壤碳库对高强度采伐干扰敏感，采伐强度决定土壤碳密度是否发生变化。 （3）估算露水河林业局林业用地在1987、1995、2003年和潜在的森林生态系统碳储量分别为29.58×106 t、27.55×106 t、30.46×106 t和38.75×106 t。

长白山阔叶红松林生态系统健康评估理论与实践

生态系统健康（学）是当前生态系统学和环境管理的研究热点，生态系统健康评估是生态系统健康（学）的主要研究内容之一。好的可操作的评估理念和技术手段是推动生态系统健康（学）研究的重要因素和内在要求。不合理的林业生产是全球森林生态系统健康程度降低的最主要因素之一。人为强度砍伐干扰直接导致林区森林面积减少、生物多样性严重破坏、资源短缺危机、森林生态系统健康程度急剧下降、提供的生命支持系统功能显著减弱。本文以长白山林区典型森林生态系统一阔叶红松林为例，提出并运用了相关森林生态系统健康评估理论和技术，进行了定量的评估研究。以期为区域森林生态系统健康管理和生态系统健康学的发展提供借鉴。研究表明：（1）通过建立恰当的概念，选择合适的原始阔叶红松林作为生态系统健康基准是可行的，并且是进行健康评估的基础；（2）干扰造成群落垂直结构、水平结构、生物多样性、生物量等具体指标的变化，导致阔叶红松林生态系统健康现状降低；（3）群落尺度上，指示物种和指示类群等生态指示法是评估生态系统健康的有效方法；（4）健康的森林生态系统是一个模式生态系统集，利用健康距离和指标体系可以在群落尺度或区域等更大尺度上对森林生态系统结构、功能、活力等多个特征进行单独或综合性健康评估；（5）综合构成指数也是一个森林生态系统健康评估的有效的方法，它在揭示森林生态系统群落结构动态和演替发展趋势上具有优越性；（6）砍伐干扰对原始森林生态系统健康具有严重破坏性。

涡动相关法估算森林生产力及与测树学方法的比较

大气中CO2、CH4和其它温室气体浓度升高导致的全球气候变化引起了人们对全球碳循环和碳收支的关注，植被与大气间CO2通量的长期测定能够加深对陆地生态系统在全球碳循环作用的科学理解。本文以我国北方典型的温带植被类型长白山阔叶红松林为研究对象，利用观测塔上的涡动相关系统对长白山阔仆卜红松林进行长期的CO2通量监测，并分析CO2通量的周年动态，估算森林净生态系统生产力；同时基于测树学方法，进行群落调查，根据已有的经验公式，估算森林净生态系统生产力，综合评价长白山阔汗卜红松林碳收支，为森林碳收支的研究提供基础。主要结论有：（1）FSAM模型的分析结果表明，观测塔上40m高度的涡动相关仪器测量的信息中，76％来自于西北至西南方相对均质的阔叶红松原始林，其中footprint最大的源区在塔西南方100m-400m范围内。因此，森林群落调查选择在此区内进行，使得涡动相关法和测树学方法估算的生产力具有可比性。（2）2003-2004年碳通量季节变化趋势基本一致，从年初到4月上旬该森林生态系统保持较弱的正的碳通量（释放CO2），5月开始表现为净的碳吸收，且吸收量迅速增加，到6月达到最大值，然后又逐渐减小；9月末到10月末随着生长季的结束，净生态系统COZ交换（NEE）开始由负转为正，11-12月NEE为正，生态系统以呼吸为主。净生态系统COZ交换的年累计量表明长白山阔叶红松林为明显的碳汇，2003年和2004年净生态系统生产力NEP分别为-217±75gcm-2a-1和-190±85gcm-2a-1，相当于-2.17±0.75tCha-1a-1和-1.90±0.85tCha-1a-1。（3）根据经验公式和材积法得到阔汗卜红松林的生物量在343．9-362．3tha－l之间，应用两种方法得到2003一2004年群落的净初级生产力在10.22-10.40tCha-1a-1之间，净生态系统生产力在2.50±1.12tCha-1a-1-2.68±1.20tCha-1a-1之间。（4）测树学方法与涡动相关法测得的净生态系统生产力略有差异，但在误差有效范围内基本一致。