复杂背景下的视频前景检测方法研究

采用计算机视觉领域的相关算法,研究了如何有效地从监控视频中提取前景目标的问题.在应用中先将视频转换成逐帧图片并依据图片每一像素的灰度值将图形以二维矩阵形式数值化.在此基础上,采用帧差法、背景差法、混合高斯模型、Harris算法与FREAK算法等多种算法,利用Matlab和Python等工具建立不同背景下的相应模型.成功地从包含动态背景、镜头晃动等不同情境的监控视频中检测并提取出前景目标.同时依据不同前景目标的各维度特征信息,利用主成分分析、k-means聚类等机器学习方法构建了判断异常事件发生的算法模型,取得了一定效果.     

多场景下复杂监控视频的前景目标提取

主要针对第十四届中国研究生数学建模竞赛D题基于监控视频的前景目标提取问题展开研究.首先建立经典模型,通过对经典模型的不足之处提出针对性的优化而建立起来的优化模型:核密度优化模型,ViBe优化模型,深度卷积神经网络优化模型.上述优化模型对监控视频前景目标提取的平均F1分数为0.83.最终提出加权模型,对视频显著目标进行提取得到包含显著前景目标的帧标号.实验结果表明,加权模型能有效提取显著前景目标,并且能成功检测出视频中的单帧.     

基于颜色和方向特征区域对比度的显著性检测模型

视觉显著性检测是很多计算机视觉任务的重要步骤,提出了一种基于颜色、方向特征和空间位置关系相结合的区域对比显著性检测算法.首先用基于图论的算法将图像分割成若干区域,结合区域间颜色特征和空间对比度计算出颜色显著图.同时采用基于纹理特征的算法分割图像,通过方向特征和空间对比度得到方向显著图.最后将二者结合得到最终显著图.在国际现有公开测试集上进行仿真实验,并与其它显著性检测方法进行对比,检测结果更加准确、合理,证明此算法切实可行.     

基于GBMTS算法的不平衡数据分类研究

解决不平衡数据分类问题,在现实中有着深远的意义。马田系统利用单一的正常类别构建基准空间和测量基准尺度,并由此建立数据分类模型,十分适合不平衡数据分类问题的处理。本文以传统马田系统方法为基础,结合信噪比及F-value、G-mean等分类精度,建立了基于遗传算法的基准空间优化模型,同时运用Bagging集成化算法,构造了改进马田系统模型算法GBMTS。通过对不同分类方法及相关数据集的实验分析,表明:GBMTS算法较其他分类算法,更能够有效的处理不平衡数据的分类问题。     

基于两阶段启发式算法的物流配送选址-路径问题研究

为了解决配送中心选址与带时间窗的多中心车辆路径优化组合决策问题,利用双层规划法建立了配送中心选址与车辆路径安排的多目标整数规划模型,针对该模型的特点,采用两阶段启发式算法进行了求解。首先,通过基于聚集度的启发式算法对客户进行分类,确定了备选配送中心的服务范围;然后,基于双层规划法,以配送中心选址成本最小作为上层规划目标,以车辆配送成本最小作为下层规划目标,建立了多目标整数规划模型;最后,利用改进的蚁群算法进行了求解。通过分析实例数据和Barreto Benchmark算例的实验结果,验证了该模型的有效性和可行性。     

基于PCA-GA-SVM的火成岩分类方法研究

在地质科学中,正确的岩石分类有助于研究岩石的成因、形成条件、演化过程和工程设计等.由于地质条件的多样性、变异性及复杂性,人们很难对岩石样本进行准确的分类.通过主成分分析法(PCA)从影响火成岩分类的众多氧化物评价指标中提取出主成分,用遗传算法(GA)优化支持向量机参数,并采用支持向量机方法(SVM)对实际火成岩公开数据进行训练,建立了火成岩岩石分类的PCA-GA-SVM模型,同时结合火成岩实际数据将预测结果和基于Levenberg-Marquardt算法改进的BP神经网络模型(LM-BP)的预测结果做了比较.结果表明:基于PCA-GA-SVM模型得到的火成岩分类预测结果精度较LM-BP神经网络有很大的提高,与实际分类相符,有广泛的应用前景.     

基于D-S证据理论的杂草图像边缘检测的研究

为了解决杂草图像边缘检测的不确定性问题,构造出图像边缘的邻域一致性、方向性和结构性三种信息测度统计,利用D-S证据理论对三种测度进行融合来实现分割后杂草图像的边缘检测,实验表明,此算法能够有效的降低噪声的影响,准确的提取出杂草边缘.     

基于Vibe算法的视频前景目标提取

前景目标提取是监控视频领域的关键性技术,对于静态背景监控的前景提取,采用Vibe算法建立背景模型,应用背景差分法得到前景目标.当监控视频中包含动态背景时,在静态背景提取的基础上引入形态学处理和中值滤波处理优化前景目标提取结果.对于摄像头抖动下所拍摄的监控视频,首先进行帧间特征角点的匹配,应用仿射变换将所有帧调整至同一角度,再将调整后的每一帧图片逐一补齐黑边,最后应用非抖动下的前景提取方法即可得到前景目标.对于含有显著前景帧号的提取,采用比例分析法简单高效地取得较优的结果.     

小鼠视觉感受区电位信号(LFP)与视觉刺激之间关系研究

针对小鼠脑电波与行为之间关系研究中涉及弱脑电波信号处理、成分分离及提取、脑电波与对应行为分析等问题,首先处理了脑电波信号,根据呼吸机理建立了"呼"、"吸"动作电位触发下的脑电波模型.其次,通过独立分量分析方法获得睡眠状态小鼠视觉感受区LFP信号模型,采用区间估计和Rayleigh熵检验结合讨论了提取的相关脑电波与呼吸信号的锁相性,分析了LFP信号的周期性,建立了多元线性回归模型,用最小二乘法从理论上讨论了其线性关系,并检验了所建二次模型周期变化的相关性.然后,通过平均经验模态分解及基于互相关性的伪分量检验算法,改善了模态混叠,建立了完整的脑电波信号分离模型.最后,分析图形因素的图像特征,通过提取视觉刺激曲线、功率谱相关频段,采用叠加平均信号法,以LFP节律随视觉信号锁相程度判断不同视觉刺激引起的LFP成分变化.     

车牌识别机器学习方法的数学建模与实现

首先,针对不同光照、复杂背景和投影失真的车牌图像建立基于Adaboost算法和改进Haar特征的车牌检测模型;然后,运用Radon变换进行车牌校正,并结合3次B样条小波变换和识别反馈模型对字符进行粗和精分割;最后,根据汉字和数字字母的不同结构特征,采用不同的算法提取特征,特别是针对车牌字符特点,训练汉字、字母和数字字母3种神经网络模型用于建立字符识别模型.实验结果表明该模型是实用的.     

统计模型在中文文本挖掘中的应用

本文讨论了中文文本挖掘的三个问题:分词、关键词提取和文本分类。对分词问题,介绍了基于层叠隐马尔可夫模型的ICTCLAS分词法,以及将词与词之间的分隔视为缺失数据并用EM算法求解的WDM方法;对关键词提取问题,提出了贝叶斯因子法,并介绍了使用稀疏回归的CCS方法;对文本分类问题,介绍了根据关键词频率建立分类器的方法,以及先建立主题模型再根据主题概率建立分类器的方法。本文通过两组文本数据对上述方法进行比较,并给出使用建议。     

带有配额的在线Nomadic旅行商问题

由于自然灾害的频繁发生,灾后的应急物资车辆调度受到了社会的广泛重视,而应急车辆尽快地将应急物资送到受灾点显得尤为重要。针对应急车辆装载物资能力有限和应急车辆不必返回出发点的情形,提出了带有配额的在线Nomadic旅行商问题。分析了该问题在正半轴和一般网络上的下界,针对受灾点仅在正半轴上的情形设计了WTAIB算法,针对受灾点在一般网络上设计了WSB算法,并进一步分析了两个算法的竞争性能。     

A primal-dual approximation algorithm for a two depot heterogeneous traveling salesman problem

Surveillance applications require a collection of heterogeneous vehicles to visit a set of targets. We consider a fundamental routing problem that arises in these applications involving two vehicles. Specifically, we consider a routing problem where there are two heterogeneous vehicles that start from distinct initial locations and a set of targets. The objective is to find a tour for each vehicle such that each of the targets is visited at least once by a vehicle and the sum of the distances traveled by the vehicles is minimal. We consider an important special case of this routing problem where the travel costs satisfy the triangle inequality and the following monotonicity property: the first vehicle’s cost of traveling between any two targets is at most equal to the second vehicle’s cost of traveling between the same targets. We present a primal-dual algorithm for this case that provides an approximation ratio of 2.     

Emergency Logistics Planning in Natural Disasters

Logistics planning in emergency situations involves dispatching commodities (e.g., medical materials and personnel, specialised rescue equipment and rescue teams, food, etc.) to distribution centres in affected areas as soon as possible so that relief operations are accelerated. In this study, a planning model that is to be integrated into a natural disaster logistics Decision Support System is developed. The model addresses the dynamic time-dependent transportation problem that needs to be solved repetitively at given time intervals during ongoing aid delivery. The model regenerates plans incorporating new requests for aid materials, new supplies and transportation means that become available during the current planning time horizon. The plan indicates the optimal mixed pick up and delivery schedules for vehicles within the considered planning time horizon as well as the optimal quantities and types of loads picked up and delivered on these routes. In emergency logistics context, supply is available in limited quantities at the current time period and on specified future dates. Commodity demand is known with certainty at the current date, but can be forecasted for future dates. Unlike commercial environments, vehicles do not have to return to depots, because the next time the plan is re-generated, a node receiving commodities may become a depot or a former depot may have no supplies at all. As a result, there are no closed loop tours, and vehicles wait at their last stop until they receive the next order from the logistics coordination centre. Hence, dispatch orders for vehicles consist of sets of “broken” routes that are generated in response to time-dependent supply/demand. The mathematical model describes a setting that is considerably different than the conventional vehicle routing problem. In fact, the problem is a hybrid that integrates the multi-commodity network flow problem and the vehicle routing problem. In this setting, vehicles are also treated as commodities. The model is readily decomposed into two multi-commodity network flow problems, the first one being linear (for conventional commodities) and the second integer (for vehicle flows). In the solution approach, these sub-models are coupled with relaxed arc capacity constraints using Lagrangean relaxation. The convergence of the proposed algorithm is tested on small test instances as well as on an earthquake scenario of realistic size.     

Using Conventional Edge Detectors and Postsmoothing for Segmentation of Spotted Microarray Images

Segmentation of spotted microarray images is important in generating gene expression data. It aims to distinguish foreground pixels from background pixels for a given spot of a microarray image. Edge detection in the image processing literature is a closely related research area, because spot boundary curves separating foregrounds from backgrounds in a microarray image can be treated as edges. However, for generating gene expression data, segmentation methods for handling spotted microarray images are required to classify each pixel as either a foreground or a background pixel; most conventional edge detectors in the image processing literature do not have this classification property, because their detected edge pixels are often scattered in the whole design space and consequently the foreground or background pixels are not defined. In this article, we propose a general postsmoothing procedure for estimating spot boundary curves from the detected edge pixels of conventional edge detectors, such that these conventional edge detectors together with the proposed postsmoothing procedure can be used for segmentation of spotted microarray images. Numerical studies show that this proposal works well in applications.

Datasets and computer code are available in the online supplements.     

An approximation algorithm for the pickup and delivery vehicle routing problem on trees

This paper presents an approximation algorithm for a vehicle routing problem on a tree-shaped network with a single depot where there are two types of demands, pickup demand and delivery demand. Customers are located on nodes of the tree, and each customer has a positive demand of pickup and/or delivery.Demands of customers are served by a fleet of identical vehicles with unit capacity. Each vehicle can serve pickup and delivery demands. It is assumed that the demand of a customer is splittable, i.e., it can be served by more than one vehicle. The problem we are concerned with in this paper asks to find a set of tours of the vehicles with minimum total lengths. In each tour, a vehicle begins at the depot with certain amount of goods for delivery, visits a subset of the customers in order to deliver and pick up goods and returns to the depot. At any time during the tour, a vehicle must always satisfy the capacity constraint, i.e., at any time the sum of goods to be delivered and that of goods that have been picked up is not allowed to exceed the vehicle capacity. We propose a 2-approximation algorithm for the problem.     

Collision Avoidance in Bi-Directional AGV Systems

The use of a bi-directional, as opposed to a one-way, transportation system, can increase the operational efficiency of automated guided vehicles in both manufacturing and warehousing environments. However, the full potential will only be realized if the routeing algorithms pay particular attention to the problem of collision avoidance. This paper is concerned with the development of such an algorithm. The method is based on the use of delays and deviations to avoid collisions that could arise when using the shortest routes for each vehicle, and is applicable to any transportation system which can be represented as a network with known travel times.     

成品油配送多车舱车辆指派及路径优化问题研究

针对成品油配送中多车型、多车舱的车辆优化调度难题,综合考虑多车型车辆指派、多车舱车辆装载及路径安排等决策,以派车成本与油耗成本之和的总成本最小为目标,建立了多车型多车舱的车辆优化调度模型。为降低模型求解的复杂性,本文提出一种基于C-W节约算法的“需求拆分→合并装载”的车辆装载策略,并综合利用Relocate和Exchange算子进行并行邻域搜索改进,获得优化的成品油配送方案。最后,通过算例验证了本文提出的模型与算法用于求解大规模成品油配送问题的有效性。并通过数据实验揭示了以下规律:1)多车舱车辆相对于单车舱车辆在运营成本上具有优越性;2)大型车辆适合远距离配送,小型车辆适合近距离配送;3)多车型车辆混合配送相对于单车型车辆配送在运营成本上具有优越性。这些规律可为成品油配送公司的车辆配置提供决策参考。     

A branch-and-cut algorithm for the multi-compartment vehicle routing problem with flexible compartment sizes

Multi-compartment vehicle routing problems arise in a variety of problem settings in which different product types have to be transported separated from each other. In this paper, a problem variant which occurs in the context of glass waste recycling is considered. In this problem, a set of locations exists, each of which offering a number of containers for the collection of different types of glass waste (e.g. colorless, green, brown glass). In order to pick up the contents from the containers, a fleet of homogeneous disposal vehicles is available. Individually for each disposal vehicle, the capacity can be discretely separated into a limited number of compartments to which different glass waste types are assigned. The objective of the problem is to minimize the total distance to be travelled by the disposal vehicles. For solving this problem to optimality, a branch-and-cut algorithm has been developed and implemented. Extensive numerical experiments have been conducted in order to evaluate the algorithm and to gain insights into the problem structure. The corresponding results show that the algorithm is able to solve instances with up to 50 locations to optimality and that it reduces the computing time by 87% compared to instances from the literature. Additional experiments give managerial insights into the use of different variants of compartments with flexible sizes.

     

应急救援物资紧缺的配送车辆路径选择研究

针对应急救援物资总量紧缺不能全部满足各点需求量的实际情形,以单个需求点最大缺货量最小为目标,建立基于单个配送中心、车辆数目有限和带时间窗的应急救援物资配送车辆路径选择模型,并讨论了三种不同缺货情形下模型的求解。对于车辆按最短路径行驶也无法在救援时间内到达导致需求点缺货的情形,删除这些需求点的需求量后,如果剩余需求点的需求量不大于物资总量且车辆足够多,可转化为车辆最短路径问题求解;对于物资紧缺无法满足所有需求且车辆足够多的情形,设计时间复杂性为O(mn²)的精确算法A^*求解,其中m和n分别为车辆数和需求点数;对于物资紧缺且车辆数目不足无法把全部物资送达需求点的情形,设计时间复杂性为O(n²)的近似算法GA^*求解,并分析了算法GA^*的近似比。最后结合云南彝良县地震灾区局部路网进行实例分析,验证模型和算法的有效性。