NUMA结构的高效实时稳定的垃圾回收算法

针对非一致性内存访问架构(NUMA)在垃圾回收(GC)过程中存在大量的远程内存读写导致GC性能降低的问题,对GC过程的各个阶段进行分析与研究,提出了一种基于NUMA结构的高效实时稳定的GC算法。该算法首先基于NUMA结构改进传统分代GC机制的堆空间布局,然后通过控制GC过程中扫描活跃对象阶段的初始根对象选取、动态负载均衡阶段截取任务队列的选取以及复制活跃对象阶段对象复制位置的选取,大大减少GC过程中的远程访问次数。这种改进的GC机制对所有NUMA结构具有通用性。以Godson-3处理器的NUMA平台为例进行的实验结果显示,优化的GC机制极大地缩短了GC的时间,而且提高了应用程序的性能以及稳定性。在SPECjvm2008测试中,GC时间平均缩短了14.6%(GC总时间缩短4.1%~41.58%),应用程序的性能平均提升了4.68%(最高提升17.8%),应用程序的性能稳定性提升了76.2%。     

面向目标检测的稀疏表示方法研究进展

目标检测作为图像理解的一个基础而重要的课题深受国内外学者的重视,在军事和民用中具有广泛应用.应用背景的多样性和复杂性使得传统目标检测算法难以克服复杂背景、噪声干扰、光照变化以及非刚体形变、遮挡、弱特征、尺度、视角和姿态变化等因素的影响.近些年来发展起来的稀疏表示方法为图像处理及目标检测研究提供了新的思路,本文概述了稀疏表示基本概念和理论研究进展,综述了稀疏表示方法在目标特征学习、目标分类器和滤波器设计以及多源信息融合目标检测等目标检测领域中的国内外重要研究进展,并展望了稀疏表示方法在目标检测领域的发展方向.     

Influence of trace Ca addition on texture and stretch formability of AM50 magnesium alloy sheet

The AM50, AM50–0.1Ca, AM50–0.3Ca and AM50–0.5Ca (wt.%) alloys were hot-rolled and their mechanical properties were determined for the purpose of investigating the effect of trace Ca addition on the texture and stretch formability of AM50 alloy. The results show that the addition of trace Ca can effectively modify the basal texture, which is characterized by the split of basal poles deviated from the normal direction (ND) after the hot rolling, while a broad spread of the basal planes toward the transverse direction (TD) after the annealing. Such change of the basal texture is related to the prior formation of massive compression twins and the decrease of the c/a ratio. Erichsen value increases from 2.25 to 4.21 mm with the increase of Ca content. The enhancement of stretch formability is ascribed to the weakened basal texture, which results in the increase of n-value and the decrease of r-value.     

Short‐term effects in a reduced flow stretch: The case of the Antas River in South Brazil

The building of adduction channels (penstocks) that conduct water from reservoirs to turbines, which are located kilometres from the dam, is becoming common, optimizing the electricity generation in small dams. This design creates a river stretch with reduced discharge between the dam and the powerhouse. This study evaluates the short‐term impacts of the below‐dam decrease in river flow on fish assemblages. Samples were collected in the reduced flow stretch of the Castro Alves Hydropower Plant (Antas River, Rio Grande do Sul, Brazil) before the reservoir started operating (January 2008; mean discharge of 103.7 m³/s) and immediately after operation began (March 2008; mean discharge of 12.4 m³/s). Sampling was conducted in distinct habitats of the reduced flow stretch (slow waters—gillnets, sand beaches—seining nets, structured littoral—electrofishing, and fast waters—cast nets) with a strongly standardized effort. The attributes of the fish assemblages were not negatively affected by the flow reduction in any habitat sampled. However, distinct changes in the spatial structure were observed considering the different types of habitat predominantly used by the species, which represents an entire reorganization of the fish assemblages in the short term. It is fundamental that these short‐term aspects be considered in the licensing of hydropower plants in addition to the long‐term changes.     

无人机影像目标检测技术研究进展

无人机影像目标检测技术是无人机影像智能化分析中的关键技术，开展无人机影像目标检测技术的研究有着广阔的应用前景和价值。介绍了无人机影像中目标检测技术的发展历程，简述了无人机影像目标检测技术在森林防火、农业信息采集、电力线路检测、路桥损害检测以及军事侦察等5种特定领域的应用情况，总结归纳了目标检测性能评价时常用的数据集和标准，并对未来无人机影像目标检测技术的发展态势进行了展望。     

A framework for cloned vehicle detection

Rampant cloned vehicle offenses have caused great damage to transportation management as well as public safety and even the world economy. It necessitates an efficient detection mechanism to identify the vehicles with fake license plates accurately, and further explore the motives through discerning the behaviors of cloned vehicles. The ubiquitous inspection spots that deployed in the city have been collecting moving information of passing vehicles, which opens up a new opportunity for cloned vehicle detection. Existing detection methods cannot detect the cloned vehicle effectively due to that they use the fixed speed threshold. In this paper, we propose a two-phase framework, called CVDF, to detect cloned vehicles and discriminate behavior patterns of vehicles that use the same plate number. In the detection phase, cloned vehicles are identified based on speed thresholds extracted from historical trajectory and behavior abnormality analysis within the local neighborhood. In the behavior analysis phase, consider the traces of vehicles that uses the same license plate will be mixed together, we aim to differentiate the trajectories through matching degreebased clustering and then extract frequent temporal behavior patterns. The experimental results on the real-world data show that CVDF framework has high detection precision and could reveal cloned vehicles’ behavior effectively. Our proposal provides a scientific basis for traffic management authority to solve the crime of cloned vehicle.     

基于AECOsimBD建立建筑三维信息模型中对象定义的方法

AECOsimBD是Bentley BIM解决方案的核心产品,它的作用是建立建筑专业的三维信息模型。而对象是三维信息建模的基本要素,文章介绍了基于AECOsimBD对象类型和对象样式的三维信息模型对象的定义方法。     

Safe system approach to reducing serious injury risk in motorcyclist collisions with fixed hazards

Collisions with fixed objects in the roadway environment account for a substantial proportion of motorcyclist fatalities. Many studies have identified individual roadway environment and/or motorcyclist characteristics that are associated with the severity of the injury outcome, including the presence of roadside barriers, helmet use, alcohol use and speeding. However, no studies have reported the cumulative benefit of such characteristics on motorcycling safety. The safe system approach recognises that the system must work as a whole to reduce the net injury risk to road users to an acceptable level, including the four system cornerstone areas of roadways, speeds, vehicles and people. The aim of the present paper is to consider these cornerstone areas concomitantly, and quantitatively assess the serious injury risk of motorcyclists in fixed object collisions using this holistic approach. A total of 1006 Australian and 15,727 (weighted) United States motorcyclist-fixed object collisions were collected retrospectively, and the serious injury risks associated with roadside barriers, helmet use, alcohol use and speeding were assessed both individually and concomitantly. The results indicate that if safety efforts are made in each of the safe system cornerstone areas, the combined effect is to substantially reduce the serious injury risk of fixed hazards to motorcyclists. The holistic approach is shown to reduce the serious injury risk considerably more than each of the safety efforts considered individually. These results promote the use of a safe system approach to motorcycling safety.     

基于双重金字塔网络的视频目标分割方法

针对复杂视频场景中难以分割特定目标的问题，提出一种基于双重金字塔网络（DPN）的视频目标分割方法。首先，通过调制网络的单向传递让分割模型适应特定目标的外观。具体而言，从给定目标的视觉和空间信息中学习一种调制器，并通过调制器调节分割网络的中间层以适应特定目标的外观变化。然后，通过基于不同区域的上下文聚合的方法，在分割网络的最后一层中聚合全局上下文信息。最后，通过横向连接的自左而右结构，在所有尺度中构建高阶语义特征图。所提出的视频目标分割方法是一个可以端到端训练的分割网络。大量实验结果表明，所提方法在DAVIS2016数据集上的性能与较先进的使用在线微调的方法相比，可达到相竞争的结果，且在DAVIS2017数据集上性能较优。