基于自动终止准则改进的kd-tree粒子近邻搜索研究

对于大规模运动模拟问题而言,近邻点的搜索效率将对整体的运算效率产生显著影响。本文基于关联性分析建立kd-tree的最大深度dmax与粒子总数N的自适应关系式,提出了kd-tree自动终止准则,即ATC-kd-tree,同时还考虑了叶子节点大小阈值n_(0)对近邻搜索效率的影响。试验表明,ATC-kd-tree具有更高的近邻搜索效率,相较于不使用自动终止准则的kd-tree搜索效率最高提升46%,且适用性更强,可求解不同N值的近邻搜索问题,解决了粒子总数N发生改变时需要再次率定最大深度dmax的问题。同时,本文还提出了网格搜索法组合坐标下降法的两步参数优化算法GSCD法。通过2维阿米巴虫形状的参数优化试验发现,GSCD法可更为快速地率定ATC-kd-tree的可变参数,其优化效率比网格搜索法最高提升了205%,相较于改进网格搜索法最高提升了90%。研究结果表明,ATC-kd-tree和GSCD法不仅提高了近邻搜索的效率,也为复杂运动中近邻粒子搜索问题提供了一种更为高效的解决方案,能够显著降低计算资源的消耗,进一步提升模拟的精度和效率。

Phase-field simulation of dendritic solidification using a full threaded tree with adaptive meshing

Simulation of the microstructure evolution during solidifi cation is greatly benefi cial to the control of solidifi cation microstructures. A phase-fi eld method based on the full threaded tree(FTT) for the simulation of casting solidifi cation microstructure was proposed in this paper, and the structure of the full threaded tree and the mesh refi nement method was discussed. During dendritic growth in solidifi cation, the mesh for simulation is adaptively refi ned at the liquid-solid interface, and coarsened in other areas. The numerical results of a threedimension dendrite growth indicate that the phase-fi eld method based on FTT is suitable for microstructure simulation. Most importantly, the FTT method can increase the spatial and temporal resolutions beyond the limits imposed by the available hardware compared with the conventional uniform mesh. At the simulation time of 0.03 s in this study, the computer memory used for computation is no more than 10 MB with the FTT method, while it is about 50 MB with the uniform mesh method. In addition, the proposed FTT method is more effi cient in computation time when compared with the uniform mesh method. It would take about 20 h for the uniform mesh method, while only 2 h for the FTT method for computation when the solidifi cation time is 0.17 s in this study.

Analysis of Cardiotocogram Data for Fetal Distress Determination by Decision Tree Based Adaptive Boosting Approach

Cardiotocography is one of the most widely used technique for recording changes in fetal heart rate (FHR) and uterine contractions. Assessing cardiotocography is crucial in that it leads to iden- tifying fetuses which suffer from lack of oxygen, i.e. hypoxia. This situation is defined as fetal dis- tress and requires fetal intervention in order to prevent fetus death or other neurological disease caused by hypoxia. In this study a computer-based approach for analyzing cardiotocogram in- cluding diagnostic features for discriminating a pathologic fetus. In order to achieve this aim adaptive boosting ensemble of decision trees and various other machine learning algorithms are employed.

Research on the adaptive hybrid search tree anti-collision algorithm in RFID system

Due to more tag-collisions result in failed transmissions,tag anti-collision is a very vital issue in the radio frequency identification(RFID) system.However,so far decreases in communication time and increases in throughput are very limited.In order to solve these problems,this paper presents a novel tag anti-collision scheme,namely adaptive hybrid search tree(AHST),by combining two algorithms of the adaptive binary-tree disassembly(ABD) and the combination query tree(CQT),in which ABD has superior tag identification velocity and CQT has optimum performance in system throughput and search timeslots.From the theoretical analysis and numerical simulations,the proposed algorithm can colligate the advantages of above algorithms,improve the system throughput and reduce the searching timeslots dramatically.

Salience adaptive morphological structuring element construction method based on minimum spanning tree

Classical mathematical morphology operations use a fixed size and shape structuring element to process the whole image.Due to the diversity of image content and the complexity of target structure,for processed image,its shape may be changed and part of the information may be lost.Therefore,we propose a method for constructing salience adaptive morphological structuring elements based on minimum spanning tree(MST).First,the gradient image of the input image is calculated,the edge image is obtained by non-maximum suppression(NMS)of the gradient image,and then chamfer distance transformation is performed on the edge image to obtain a salience map(SM).Second,the radius of structuring element is determined by calculating the maximum and minimum values of SM and then the minimum spanning tree is calculated on the SM.Finally,the radius is used to construct a structuring element whose shape and size adaptively change with the local features of the input image.In addition,the basic morphological operators such as erosion,dilation,opening and closing are redefined using the adaptive structuring elements and then compared with the classical morphological operators.The simulation results show that the proposed method can make full use of the local features of the image and has better processing results in image structure preservation and image filtering.

MINIMUM ATTRIBUTE CO-REDUCTION ALGORITHM BASED ON MULTILEVEL EVOLUTIONARY TREE WITH SELF-ADAPTIVE SUBPOPULATIONS

Attribute reduction is an important process in rough set theory.Finding minimum attribute reduction has been proven to help the user-oriented make better knowledge discovery in some cases.In this paper,an efficient minimum attribute reduction algorithm is proposed based on the multilevel evolutionary tree with self-adaptive subpopulations.A model of multilevel evolutionary tree with self-adaptive subpopulations is constructed,and interacting attribute sets are better decomposed into subsets by the self-adaptive mechanism of elitist populations.Moreover it can self-adapt the subpopulation sizes according to the historical performance record so that interacting attribute decision variables are captured into the same grouped subpopulation,which will be extended to better performance in both quality of solution and competitive computation complexity for minimum attribute reduction.The conducted experiments show the proposed algorithm is better on both efficiency and accuracy of minimum attribute reduction than some representative algorithms.Finally the proposed algorithm is applied to magnetic resonance image(MRI)segmentation,and its stronger applicability is further demonstrated by the effective and robust segmentation results.

Studies on adaptability of Populus bolleana and other Poplar trees in northwest Shanxi Province

Through five years (1996-2000) experiment, the growth characteristics and resistance to insect and cold of the seven poplar varieties including P. po pularis, P. opera, P. pseudo-simonii, P. beijinensis, P. bolleana, P. ( eurameri cana and P. simonii, which were planted in loess hilly land in northwest Shanxi Province, were tested under different planting densities and site conditions. A randomized block design and three times repetition were adopted in the test. Bas ed on the test results in arid and cold area of the northwest Shanxi, the hybrid poplar trees taking Populus cathaysna as their female parent could be selected for afforestation, such as P. popularis and P. opera, and the growing space of s ingle plant should be larger than 20 m2. P. bolleana as an introduced tree speci es grows well in this area, but its growth is likely affected by ground water co ndition. On the site with relatively abundant soil water, P. bolleana usually su ffer from frost crack at its trunk base. As a result, P. bolleara is suited to p lanting in the site where is short of soil water. P. x euramericana is not suita ble for large-scale afforestation, especially in arid slopes and ridges of loess hilly land .

SVM-DT-Based Adaptive and Collaborative Intrusion Detection

As a primary defense technique, intrusion detection becomes more and more significant since the security of the networks is one of the most critical issues in the world. We present an adaptive collaboration intrusion detection method to improve the safety of a network. A self-adaptive and collaborative intrusion detection model is built by applying the Environmentsclasses, agents, roles, groups, and objects(E-CARGO) model. The objects, roles, agents, and groups are designed by using decision trees(DTs) and support vector machines(SVMs), and adaptive scheduling mechanisms are set up. The KDD CUP 1999 data set is used to verify the effectiveness of the method. The experimental results demonstrate the feasibility and efficiency of the proposed collaborative and adaptive intrusion detection method. Also, the proposed method is shown to be more predominant than the methods that use a set of single type support vector machine(SVM) in terms of detection precision rate and recall rate.

FIRE CYCLE AND FIRE ADAPTATION OF MAIN TREE SPECIES IN DAXING'ANLING FOREST REGION, CHINA

Based on the fire statistics, the Daxing'anling forest area were classified into three fire cycle regions: northern coniferous virgin forest region with a fire cycle of 110-120 years, middle conifer--broad-leaved mixed forest region with a fire cycle of 30-40 years, and southem broad-leaved secontw forest region with a fire cycle of 15-20. The percentage of conifers and broad-leaved trees, for est age and natural mature period of main tree species in different fire cycle regions were discussed in concem with fire occurrence. The characteristics of fire adaptation and fire resisboce of main tree species, such as sexual and asexual reproduction, were discussed and evaluation of the synthetical fire adaptation was made.

Agro-Ecological Adaptation and Participatory Evaluation of Multipurpose Tree and Shrub Legumes in Mid Altitudes of Sud-Kivu, D. R. Congo

Livestock is traditionally managed in mixed crop-livestock production systems in Sud-Kivu, D. R. Congo. Currently, livestock production is facing numerous constraints due to wars and insecurity in the country, with looting of animals, demographic pressure on natural resources, and lack of extension services. Multipurpose trees and shrubs with adequate forage quality could help overcome dry-season feed shortage. The objective of this research was to assess the agro-ecological adaptability of selected forage tree and shrub legumes combined with farmer participatory evaluation. The study was carried out at two distinct mid-altitude sites in Sud-Kivu. Trees and shrubs were planted randomly in eleven lines, with each species/accession consisting of nine plants split into three replications. Following a standardization cut after one year of growth, regular biomass harvests were performed every eight weeks during one year;Desmodium and Flemingia were cut at 0.5 m above soil surface, while Leucaena and Calliandra at 1 m. Before every harvest, plant height was measured and number of stems counted. Fresh leaf and stem biomass were weighted, sub samples dried and leaves analyzed for nutritive value. Results from agronomic evaluation show that Leucaena diversifolia ILRI 15551 and Leucaena leucocephala had the highest yield in the rainy season on fertile soil;during dry season, Leucaena diversifolia ILRI 15551 was superior. On poor soil, Flemingia macrophylla CIAT 17403 performed best in both seasons, however, its forage quality was relatively low. Forage tree and shrub legumes chosen by farmers were, in general, the same as those superior ones in agronomic evaluation.

ADAFT:SDN大规模流表的适应性深度聚合存储架构

为解决软件定义网络(SDN)数据平面中的三态内容可寻址存储器(TCAM)资源紧张问题,提出了一种基于内容表项树的SDN流表深度聚合方法,进而构建一种SDN大规模流表的适应性深度聚合存储架构ADAFT。该架构放宽了聚合表项之间的汉明距离要求,构建内容表项树聚合动作集不同的流表项,显著提高了流表聚合程度。设计了一种TCAM装载率感知的内容表项树动态限高机制,以降低流表查找开销。同时,提出了一种TCAM装载率感知的表项聚合适应性选择策略,以均衡流表聚合程度和查找开销。实验结果表明,ADAFT架构的流表压缩率明显高于现有方法,最高可达65.74%。

Learned Index和B-Tree在不同分布数据上的性能对比及优化

Learned Index是一种通过训练模型来建立输入数据和存储位置之间映射关系的索引,它能学习到数据间分布的信息,而不同的数据分布将影响模型训练准确率和模型复杂度之间的平衡。为了探索Learned Index适用的场景,使用不同分布、不同数据量的数据对它和加以优化的可更新的自适应学习索引(ALEX)进行性能测试,并与B-Tree进行对比,最终发现Learned Index构建大批量数据的索引时间比B-Tree短,读操作性能、存储空间大小有明显的优势,但写操作性能较差,因此得出Learned Index更适用于大数据情景下的在线分析处理(OLAP)数据库,用于静态数据的存储和查询操作的结论。基于B-Tree的索引结构,对初版Learned Index的结构进行了优化和调整,最终使优化后Learned Index在大批量数据的读写操作性能上有明显提高,其中读操作最高达到原版Learned Index的2倍,写操作最高达到原版的3倍。

近海复杂环境下UUV动态路径规划方法研究

为解决近海环境下水下无人航行器(unmanned underwater vehicle,UUV)的动态路径规划问题,本文提出一种结合全局和局部动态路径规划的算法。首先,本文提出一种基于自适应目标引导的快速拓展随机树算法,以增加随机树生长的方向性,并通过转向和重选策略减少无效拓展加快算法的收敛速度。接着,获得全局路径之后使用自适应子节点选取策略获取动态窗口法的子目标点,将复杂的全局动态任务规划分解为多个简单的动态路劲规划,从而防止动态窗口法陷入局部极小值。最后,通过UUV出港任务仿真实验验证了算法的有效性和实用性。

基于RRT^(*)改进的移动机器人路径规划算法

针对RRT^(*)算法在复杂环境路径规划中存在的盲目搜索、冗余节点及路径较长等问题,提出一种融合树扩展策略和采样策略的改进RRT^(*)算法(AF-RRT^(*))。通过创造父节点改进RRT^(*)扩展树的结构,缩小路径长度;引入自适应探索,增加采样导向的选择性,减少路径搜索时间,同时不会陷入局部最优陷阱;通过动态步长,减少冗余节点。仿真结果表明,AF-RRT^(*)算法在多种环境下,路径获取效率和路径质量均优于RRT^(*)和F-RRT^(*)。消融实验验证了AF-RRT^(*)算法和算法各功能模块的有效性。

基于激光雷达自适应聚类半径的树冠检测研究

为解决丘陵山地果园条件下激光雷达检测过程中面对多尺寸、多距离条件下出现的目标物体漏检、误检等欠分割和过分割问题,提出一种基于激光雷达的自适应目标聚类半径目标物体检测方法。首先,在使用激光雷达感知到周围环境的三维点云后,去除地面点云并且使用体素滤波进行降采样的预处理,在减少数据量的同时去除点云中的噪声点。其次,建立K-d tree模型进行最近邻搜索,以加速欧式聚类的进程,通过自适应确定每颗树冠的聚类半径,使欧式聚类能够得到更好的聚类效果。最后为验证算法准确性和实用性,基于果园履带车平台,采用32线激光雷达对所提算法进行实车测试。结果表明:在丘陵山地果园中该算法可准确聚类果树树冠点云,且实地目标正检率为94.41%。

基于故障树分析法的舰载武器装备电子元器件及PCB的失效研究

目的 对极端海洋环境下某型舰声呐装置中的多功能信号发生器进行失效研究。方法 使用故障树分析法作为失效分析基本方法。首先通过微焦点CT检测技术对样品进行准确的失效定位,并通过扫描电子显微镜和能谱分析样品的失效机理。然后结合应用背景,分析归纳样品失效的机理因子和影响因素。最后,针对每种失效形式提出具体的改善方案。结果 由电阻本身表面形貌、额定温度及外部环境(高温、高湿、高盐)等原因,导致含有盐雾的水汽进入电阻内部,致使电阻内部出现电化学迁移现象,电阻阻值因此明显降低,最终导致多功能信号发生器出现失效。结论 舰艇中的电子设备,如多功能信号发生器的失效往往是由设计制造及各种海洋环境因素综合作用所造成的。这就要求舰载电子设备的研制方、使用方都应重视海洋环境适应性问题,并采取一定的改善措施。

不同树龄国槐叶片性状特征及其适应策略研究

叶片性状的变异与相关关系能够反映植物的资源利用方式和适应生存策略。为探讨不同树龄国槐叶片性状特征及其变异规律,以聊城市不同树龄国槐叶片为研究对象,通过测定叶面积(leaf area,LA)、叶鲜重(leaf fresh weight,FW)、叶干重(leaf dry weight,DW)、叶长(leaf length,LL)、叶宽(leaf width,LW)、叶柄长(petiole length,PL)、叶厚度(leaf thickness,LT)、叶轴长(rachis length,RL)、叶形指数(leaf shape index,LI)、比叶面积(specific leaf area,SLA)、叶干物质含量(leaf dry matter content,LDMC)、叶组织密度(leaf tissue density,LTD)、比叶重(specific leaf weight,SLW)共13项指标,运用单因素方差分析、变异系数分析探究不同树龄叶片性状特征,经相关性分析、主成分分析探究各树龄国槐叶片的适应规律。结果表明,后备资源的DW、LL、PL极显著高于其他树龄的(P<0.01);国槐13个叶片性状的总体变异系数在8.57%~81.48%,不同树龄叶片性状中后备资源的LTD变异系数最大,为92.86%,2级古树LDMC变异系数最小,仅5.84%;LA、FW、DW、LL和LW呈极显著正相关(P<0.01),SLA与FW、DW、LTD、SLW呈极显著负相关(P<0.01),与LT呈极显著正相关(P<0.01);LDMC与FW、DW、LTD、SLW呈极显著正相关(P<0.01);后备资源叶片性状呈高LDMC、SLW,低SLA的特点,采取的是“快速投资-收益型”策略;2级古树和1级古树呈高SLA,低LDMC、SLW的叶片性状组合,采取的是“缓慢投资-收益型”策略。研究结果为制定不同龄级国槐养护管理策略提供科学依据。

基于集成学习的交通事故严重程度预测研究与应用

目前自动驾驶技术重点是关注如何主动避免碰撞,然而在面对其他交通参与者入侵而导致不可避免的碰撞事故场景时,预测车辆在不同行驶模式下的碰撞严重程度来降低事故严重程度的研究却很少。为此,提出一种双层Stacking事故严重程度预测模型。基于真实交通事故数据集NASS-CDS完成训练,模型输入为车辆传感器可感知得到的事故相关特征,输出为车内乘员最高受伤级别。在第1层中,通过实验对不同学习器组合进行训练,最终综合考虑预测性能以及耗时挑选K近邻、自适应提升树、极度梯度提升树作为基学习器;在第2层中,为降低过拟合,采用逻辑回归作为元学习器。实验结果表明,该方法准确率达到85.01%,在精确率、召回率和F1值方面优于其他个体模型和集成模型,该预测结果可作为智能车辆决策规划模块先验信息,帮助车辆做出正确的决策,减缓事故损害。最后阐述了模型在L_(2)辅助驾驶与L_(4)自动驾驶车辆中的应用,在常规车辆安全防护的基础上进一步提升车辆的安全性。

激光自适应光学系统可靠性改进效果评估

随着自适应光学技术在激光领域的发展,工程上以经典自适应光学(AO)系统为基础,增加了多种基于软件监测和硬件保护的改进措施以保证激光AO系统稳定连续出光。面对结构复杂度提升带来的可靠性挑战,如何构建系统失效模型对激光AO系统可靠性进行评估,成为影响激光AO系统发展的重要一环。本文以激光光稳净化AO系统为例,提出使用动态故障树方法对激光AO系统可靠性进行评估,根据设备间动态关系建立动态故障树(DFT),结合厂家信息、疲劳寿命试验与历史数据估计得到底事件失效率,使用二元决策图和马尔可夫模型求解得到DFT的可靠性参数。使用DFT分析增加改进措施的AO系统可靠运行时间,结果相对于基本故障树获得了十倍以上的提高。实际系统调试期间,在预计的可靠运行时间内未发生自因故障,与DFT估计结果一致。验证了应用DFT方法评估增加改进措施后的激光AO系统可靠性更准确。

自适应Tree-Mesh结构的大棚无线监测网络设计

针对大棚基地作物状态及环境信息的无线采集的需求,设计了改进的分簇Tree-Mesh混合拓扑结构无线传感器网络,并利用ZigBee实现了组网和多跳通信,以CC2530为核心设计了多传感器无线节点硬件系统,基于Z-Stack协议栈设计了有限状态机节点程序。同时,针对无线节点低功耗和网络信息低冗余的要求,设计了基于接收信号强度指示的最佳发射功率自适应机制,和基于感知数据差值的最小传输数据冗余自适应机制。试验结果表明,节点单跳和多跳通信速率典型值分别为20与0.3kb/s,采用干电池供电和直流供电的节点通信距离分别可达30和90m。仿真结果证明采用低功耗自适应机制的节点功耗降低了38.44%,可用作大棚基地的环境监测。