基于混合粒子PHD滤波的多目标视频跟踪

针对可变数目多目标视频跟踪,粒子滤波不能持续维持目标的多模态分布问题,本文提出一种混合粒子概率假设密度(PHD)滤波的多目标视频跟踪算法.该算法首先用K-means算法对粒子进行空间分布聚类,给各粒子群附加身份标签,使各粒子群分别对应混合粒子滤波的各分量,采用相互独立的各分量粒子滤波跟踪各目标,这样提高了目标状态估计的准确性,也能有效维持各目标的多模态分布.实验结果表明,该算法能有效处理新目标出现、合并、分离等多目标跟踪问题.     

基于概率分布模型的流量预报及参数动态识别

概率分布模型(PDM)在国内流域的适用性及参数在不同径流阶段的变化研究较少。本文以浙江省两典型流域为例,建立PDM流量预报对比模型,根据参数动态识别分析确定不同时间窗口下参数变化,从而推断关键影响因子。研究结果表明:(1)PDM模型在两流域模拟结果较为满意,纳什效率系数均可达到0.7以上,且低流量为主的龙泉溪流域模拟效果明显优于高流量主导的金华江流域;(2)在不同洪水事件下PDM参数(经验参数α、b和最大蓄水能力S_max)均与前期土壤湿度显著负相关,与退水斜率有关的参数b在金华江流域与蒸发量显著正相关,而在龙泉溪流域与平均降雨负相关性显著;(3)受地表以下28 cm土壤湿度变化控制的参数α在洪峰及退水阶段识别度最大。     

带雨淋水膜的非耐热防火玻璃的火灾试验

研究了水膜对防火玻璃耐热性质的影响。用标准试验炉对门窗玻璃做火灾试验，温度高达1000℃。结果表明，对玻璃施用水膜能有效地经受试验的高温，水膜的蒸发潜热被用于保护建筑物的玻璃，免于因火灾造成融熔或炸裂。并且保留其完整性和绝缘性的试验时间由6min延长到100min。值得注意的是，玻璃表面的温度维持在90℃以下可持续100min。结论：用带水膜的防火玻璃替代通常装设的门是可行的。     

Transformations of a class of non-linear multivariable control systems to block companion forms and their applications

The problem of transforming a class of multivariable non-linear systems to the controller and observer block companion forms is considered for pole assignment and observer design. Solving non-linear partial differential equations is needed for finding the transformations and the conditions for the existence of such transforms are stated.     

Analysis on repeat-buying patterns

Consumer market has several characteristics in common such as repeat-buying over the relevant time frame, a large number of customers, and a wealth of information detailing past customer purchases. Analyzing the characterizations of repeat-buying is necessary to understand and adapt to dynamics of customer behaviors for company to survive in a continuously changing environment. The aim of this paper is to develop a methodology to detect the existence of repeat-buying behavior and discover the potential period of the repeat-buying behavior. We propose a new mathematical model to capture the characteristics of repeat-buying behavior. The algorithms based on our previous works then proposed to provide a scheme to discover periodicity and trends of the purchase. Two fundamental repeat-buying types have been identified and analyzed. Any repeat-buying scenarios can be expressed as the combination of the two fundamental types. The proposed mathematical model coupled with our prior works on cyclic modeling form a systematic process to uncover the characteristics of repeat-buying phenomenon. The experiments against a domestic consumer goods company are provided. The experimental results show that the proposed model can predict likely periodic purchase more precisely than previous studies.     

Robust stabilization against non-linear time-varying uncertainties

A simple robust stability criterion for a linear multivariable system is developed to guarantee the stability of the closed-loop system subject to non-linear time-varying uncertainties. The generally parameterized stabilizing controller advanced by Youla et al. is combined with the Schur function (class S) to synthesize the robust stabilizer of feedback systems subject to these uncertainties. Both additively and multiplicatively non-linear uncertainties are considered in this paper. Two examples are given to demonstrate the validity of our results.     

Adaptive control with optimal robust tracking

The problem of optimal robust tracking in two-parameter adaptive control systems under non-linear time-varying unmodelled dynamics is examined. A new robust stability criterion is derived for analysing the robustness of adaptive control systems with non-linear time-varying model errors. Based on the concept of excess robustness and the theory of the minimum H^∞norm, a simple and feasible design algorithm is presented to synthesize a two-parameter adaptive controller which ensures that adaptive control systems can achieve the object of optimal robust tracking in the presence of non-linear time-varying unmodelled dynamics. Simulation results that demonstrate features of the two-parameter adaptive controller with optimal robust tracking in the light of the design algorithm are included.     

Non-linear algorithm for the optimal estimation of multiple localized targets

For allocating the localized target in tracking, auxiliary data and signal processors are needed to achieve the measurement in the related different localized target. An efficient recursive algorithm to track multiple targets is developed in this paper for the case of the localized surveillance system. The main effort in the estimation of the filtering and prediction filter applied in tracking targets is concentrated on the error covariance and filter gain. The advantages of this residual algorithm are the use of prestored data and the fact that the target trace has a reasonable convergence computation rate for localized target estimation and prediction.     

一种用于预防支路型失稳故障的灵敏度方法

提出了一种新的用于评估支路型失稳故障严重程度的指标，并提出了一种可用于预防支路型失稳故障的灵敏度方法。严重的支路故障会造成电力系统立即失去静态稳定性。这些失稳的支路型故障都有一个对应于支路参数空间的虚拟的静态稳定临界点。本文提出了基于这一稳定临界点的灵敏度分析方法，并将之应用于对支路型失稳故障的预防控制，取得了与负荷参数空间中稳定临界点处灵敏度方法同样的控制效果。文中通过一个600节点的实际系统算例说明本文所提方法的有效性。     

Passion Fruit Juice Concentration by Ultrafiltration and Evaporation

Passion fruit juice was pretreated with pectinase, centrifugation, and pasteurization. The resulting juice was processed by Ultrafiltration (UF) to 20°Brix. The UF permeate was concentrated by evaporation to 70°Brix and combined with the UF retentate to form the final concentrated product at 40°Brix. The pretreatments caused approximately 20% flavor losses, but accomplished more than 50% flux increases. The UF recovered many important flavor constituents, and improved the rate of evaporation. The final concentrated product contained more than 30% of juice flavor. Results of sensory evaluation indicated that flavor of the reconstituted juice was inferior to fresh juice, but the two juices were not significantly different in overall acceptance.