融合重检测机制的卷积回归网络目标跟踪算法

针对基于人工特征的背景感知相关滤波（CACF）算法在形变、运动模糊、低分辨率情形跟踪效果较差以及跟踪器遇到严重遮挡等情形容易陷入局部最优而导致跟踪失败的问题，提出一种融合重检测机制的卷积回归网络（CRN）目标跟踪算法。在训练阶段，将相关滤波作为CRN层融入进深度神经网络，使网络成为一个整体进行端到端训练；在跟踪阶段，通过残差连接融合不同网络层及其响应值，同时引入重检测机制使算法从潜在的跟踪失败中恢复，当响应值低于给定阈值时激活检测器。在数据集OTB-2013上的实验表明，所提算法在50个视频序列上精确度达到88.1%，相比原始CACF算法提高9.7个百分点，在具有形变、运动模糊等属性的视频序列上相比原始算法表现更优秀。     

Study on downward flame spread over extruded polystyrene foam in a vertical channel: Influence of opening area

Experimental methods and theoretical analysis are employed to investigate effects of channel opening area on downward flame spread characteristics of extruded polystyrene (XPS) thermal insulation materials on building facade. The average flame height first drops and then rises as dimensionless opening area (the ratio of sidewall opening area to sidewall area, ie, S^*) increases. As S^* rises, both the average and maximum temperature of the curtain wall decrease, and the decreasing of the average temperature is linear. XPS surface temperature history can be divided into four stages, ie, inapparent rising stage (preheating), significant rising stage (melting), dropping stage (pyrolysis), and rerising stage (ignition). The preheating length first rises and then drops as S^* increases. The XPS flame spreads steadily at the early period while acceleration occurs at the later period. For different opening areas, the difference in spread distance history is not apparent in the early stage while this difference is significant in the later stage. The flame spread rate (V_f) first increases and then decreases as S^* rises. A downward flame spread model for XPS in vertical channel with openings is built. The varied trend of V_f predicted using this model corresponds to the experimental result.     

高斯混合模型结合加权似然的目标跟踪算法

鉴于高斯混合模型对背景变化快时无法精确检测出目标和目标跟踪的适应性差等瑕疵，提出了基于加权似然跟踪器来改进高斯混合模型实现运动目标跟踪算法。主要引入了自适应高斯混合模型来实时检测运动目标，然后空间加权似然来进行视频中的目标定位，引入加权似然期望值来改进高斯混合模型处理视频中的多尺度、多角度变化的目标跟踪不精准问题。通过VOT 2014 dataset对比实验结果表明提出的基于加权似然跟踪（Weighted Likelihood Tracking，WLT）和改进高斯混合模型（Improved Gaussian Mixture Model，IGMM）的目标跟踪算法较传统高斯混合模型跟踪算法在跟踪的精度有较大提高。在应对多尺度、多角度变化的目标跟踪表现出了较大的优势。     

水下循迹航行器水动力学性能数值研究

为了研究低速水下循迹监测航行器的水动力学性能数值计算问题,采用FLUENT软件和SST剪切应力输运模型,通过雷诺时均N-S方程分析流速一定的情况下,取不同攻角、不同水平舵角作为来流条件,研究未安装推进器以及安装推进器且其安装位置不同时,航行器的升力系数、俯仰力矩系数、表面压力分布和流场速度的变化规律。结果表明:在未安装推进器以及推进器的安装位置不同时,随着攻角的变化,升力系数呈线性变化,俯仰力矩系数呈非线性变化;随着水平舵角的变化,升力系数和俯仰力矩系数呈线性变化。当推进器安装在航行器头部时,对航行器流场压力和流场速度变化影响最大;当安装在航行器尾部时,对二者影响最小。对于低速航行器,应尽量将推进器安装在中间靠后位置,以提高航行器的水动力性能。升力系数的试验结果与数值仿真结果之间最大相差7. 51%,阻力系数的试验结果与数值仿真结果最大相差5. 84%,均吻合较好。研究结果可以为低速水下循迹航行器的优化设计和发展提供理论参考。     

基于微流控的人卵母细胞特性检测

人的卵母细胞特性的研究对于临床医学具有重要意义,可以通过卵母细胞的大小对其活性进行研究。采用微流控平台与MATLAB算法相结合的方法,利用微流控器件的沟道对卵母细胞进行挤压,由于卵母细胞的活性不同对于挤压的反应也不同,活性越好的卵母细胞通过沟道时变形性越好,经检测卵母细胞有很好的活性;拍摄卵母细胞发生形变与卵母细胞通过沟道的视频,利用MATLAB程序进行图像处理得到卵母细胞的大小,测得卵母细胞的直径约为170μm。这种方法有别于传统方法,在一定程度上使用了算法识别,提高了卵母细胞分析统计的准确度,提升了工作效率。由于人的卵母细胞比较珍贵,只进行了初期实验研究,证明此方法确实可行。     

Enhancement of resistance to galvanic corrosion of ZE41 Mg alloy by equal channel angular pressing

The galvanic corrosion behavior of as-received and ECAPed ZE41 Mg alloy coupled with Al7075 alloy is investigated using zero resistance ammeter in three different corrosive environments, 0, 0.1, and 1 M NaCl, to mimic the conditions experienced in engineering applications. The mechanism of galvanic corrosion for the ZE41 Mg–Al7075 aluminum alloy is explained. It is observed that a robust surface film containing a composite layer of oxide/hydroxide of magnesium and aluminum is established in deionized water (0 M). However, only a single layer of magnesium oxide/hydroxide is detected in chloride-containing environments. Equal channel angular pressing (ECAP) improved the resistance to galvanic corrosion by 58% and 54% when compared with the as-cast counterparts in 0 and 1 M NaCl solution, respectively. In contrast, galvanic corrosion resistance decreased by 26% in 0.1 M NaCl after ECAP while the as-received samples evinced pits unfavorable to be used in engineering applications. ECAP is a promising method to combat galvanic corrosion encountered by ZE41 magnesium alloy used in automobiles and components of military vehicles.     

Modeling and analysis of multiple access channel capacity based on hybrid energy storage and energy harvesting

Energy harvesting (EH) has been considered as a promising technology to solve the constrained energy problem in the devices of IoT with its advantages of flexible deployment and sustainable energy supply.For multiple access channel with energy harvesting,a hybrid energy storage structure model composed by super capacitor and battery was proposed for the devices of IoT.According to the peculiarities of medium access channel and energy harvesting system,an optimized energy allocation strategy with exponential-type decline (ETD) was presented,the upper and lower bounds of the average throughput were deduced,in particular,the gap of two bounds was derived to be a constant.The channel capacity was further obtained by utilizing the relationship between the average throughput and the channel capacity.In the simulations,the effect of harvested energy,storage capacity and the number of nodes on the channel capacity were analyzed respectively.Experiment results show that compared with the conventional wireless node with single battery storage,the proposed hybrid energy storage structure can improve the harvested energy value and increase the multiple access channel capacity by using adaptive modulation scheme when transmitting the signals.     

Temporal information based GoP adaptation for linear video delivery schemes

An original wireless video transmission scheme called SoftCast has been recently proposed to deal with the issues encountered in conventional wireless video broadcasting systems (e.g. cliff effect). In this paper, we evaluate and optimize the performance of the SoftCast scheme according to the transmitted video content. Precisely, an adaptive coding mechanism based on GoP-size adaptation, which takes into account the temporal information fluctuations of the video, is proposed. This extension denoted Adaptive GoP-size mechanism based on Content and Cut detection for SoftCast (AGCC-SoftCast) significantly improves the performance of the SoftCast scheme. It consists in modifying the GoP-size according to the shot changes and the spatio-temporal characteristics of the transmitted video. When hardware capacities, such as buffer or processor performance are limited, an alternative method based only on the shot changes detection (AGCut-SoftCast) is also proposed. Improvements up to 16 dB for the PSNR and up to 0.55 for the SSIM are observed with the proposed solutions at the cut boundaries. In addition, temporal visual quality fluctuations are reduced under 1dB in average, showing the effectiveness of the proposed methods.