基于多专家区间数的多属性群决策方法

针对区间数的多专家多属性决策问题,提出了一种基于非线性规划模型的群决策方法。该方法建立如下准则:在不同对象和属性下,当某专家的估计值与所有专家估计值的均值越靠近时,则其专家权重就越大;反之就越小。基于该准则利用区间距离公式和规划模型解决了专家权重难以确定的问题。结合集成算子理论,利用区间数算术平均算子将决策矩阵集成为综合决策矩阵,再利用属性权重将其集成为综合属性值,通过二维可能度建立比较可能度矩阵,然后利用排序向量法进行排序。最后通过实例分析验证了该方法的可行性和合理性。     

基于改进SIFT特征提取的自主移动机器人环境识别方法研究

为了提高机器人在环境识别中的实时性和准确性,在分析了SIFT描述算子性能及原理的基础上,对SIFT算法在计算描述算子阶段提出自适应修改步长的改进,以及在原先的匹配方法的基础上又加入二次反向匹配,既在一定程度上提高了实时性又对关键点的消除错配起到较好的效果。仿真结果也证明了其有效性。     

An aggregation model reduction method for one‐dimensional distributed systems

A method for deriving reduced dynamic models of one‐dimensional distributed systems is presented. It inherits the concepts of the aggregated modeling method of Lévine and Rouchon originally derived for simple staged distillation models and can be applied to both spatially discrete and continuous systems. The method is based on partitioning the system into intervals of steady‐state systems, which are connected by dynamic aggregation elements. By presolving and substituting the steady‐state systems, a discrete low‐order dynamic model is obtained. A characteristic property of the aggregation method is that the original and the reduced model assume identical steady states. For spatially continuous systems, the method is an alternative to discretization methods like finite‐difference and finite‐element methods. Implementation details of the method are discussed, and the principle is illustrated on three example systems, namely a distillation column, a heat exchanger, and a fixed‐bed reactor. © 2011 American Institute of Chemical Engineers AIChE J, 2012     

脂肪族醇的酸碱离解常数(pKa)的预测研究

酸碱离解常数(pKa)在理解受体与药物反应和药物如何选择辅料上起着重要的作用。因此,在药物前期的开发中,能有效地预测pKa值可很好地降低开发成本,降低资源消耗。本文通过运用定量构效关系的方法(QSAR)对脂肪醇的数据库展开研究,以进行pKa值的预测。一共选用了253种各类描述符,包含原子键的性质的描述符48个,分子整体描述符205个,其中重点描述符为原子键的性质类描述符。采取多元线性回归的方法对数据进行研究。该预测模型能取得较好的效果,可以应用于药物的前期开发中。     

An ant colony optimisation algorithm for aggregated multicast based on minimum grouping model

The tree‐based delivery structure of the traditional Internet protocol multicast requires each on‐tree router to maintain a forwarding state for a group. This leads to a state scalability problem when large numbers of concurrent groups exist in a network. To address this state scalability problem, a novel scheme called aggregated multicast has recently been proposed, in which multiple groups are forced to share one delivery tree. In this paper, we define the aggregated multicast problem based on the minimum grouping model, and propose an ant colony optimisation algorithm. The relative fullness of the tree is defined according to the characteristics of the minimum grouping problem and is introduced as an important component in identifying the aggregation fitness function between two multicast groups. New pheromone update rules are designed based on the aggregation fitness function. To improve the convergence time of the algorithm, we use the changes (brought by each group) in the relative fullness of the current tree as the selection heuristic information. The impact of the relative fullness of the tree is analysed using the hypothesis test, and simulation results indicate that introducing relative fullness to the fitness function can significantly improve the optimisation performance of the algorithm. Compared with other heuristic algorithms, our algorithm has better optimisation performance and is more suitable for scenarios with larger bandwidth waste rates. Copyright © 2011 John Wiley & Sons, Ltd.     

基于傅立叶描述子的步态识别

提出了基于傅立叶描述子的步识别方法。用背景差方法得到运动人体的轮廊,通过步态周期分析提取步态序列的关键帧。利用傅立叶描述子处理关键帧的轮廊线序列,并进行数据维数压缩,得到匹配模板。用最近邻近法进行分类和识别。应用上述方法在Soton步态数据库上进行了实验,结果表明所提的步态识别方法具有罗高的识别性能。