Exponential stability of general tracking algorithms

Tracking and adaptation algorithms are, from a formal point of view, nonlinear systems which depend on stochastic variables in a fairly complicated way. The analysis of such algorithms is thus quite complicated. A first step is to establish the exponential stability of these systems. This is of interest in its own right and a prerequisite for the practical use of the algorithm. It is also a necessary starting point to analyze the performance in terms of tracking and adaptation because that is how close the estimated parameters are to the time-varying true ones. In this paper we establish some general conditions for the exponential stability of a wide and common class of tracking algorithms. This includes least mean squares, recursive least squares, and Kalman filter based adaptation algorithms. We show how stability of an averaged (linear and deterministic) equation and stability of the actual algorithm are linked to each other under weak conditions on the involved stochastic processes. We also give explicit conditions for exponential stability of the most common algorithms. The tracking performance of the algorithms is studied in a companion paper     

通用进化算法的收敛性分析

传统进化算法的收敛性专注于具体算法,对应的研究成果也仅仅适用于具体算法。为了研究所有进化算法的收敛性问题,提出了一种包含所有操作类型算子的通用进化算法,建立了一套概率空间用于研究算法的收敛性,所有有关算法的术语都用严格的数学语言加以定义。在概率空间中,有七个算法收敛性定理被完整地证明,其中之一找到了算法依概率收敛的充分必要条件。更为重要的是,这些定理适用所有进化算法。它建立了一个体系,用来指导进化算法的设计,从理论上判断进化算法的收敛性。     

非线性滤波算法在机动目标跟踪中的研究

介绍了3种最基本非线性滤波算法--扩展卡尔曼滤波(EKF)、无迹卡尔曼滤波(UKF)和粒子滤波(PF)算法的理论在机动目标跟踪中的应用.通过仿真试验对三者性能进行了分析比较.     

Performance analysis of three text-join algorithms

When a multidatabase system contains textual database systems (i.e., information retrieval systems), queries against the global schema of the multidatabase system may contain a new type of joins-joins between attributes of textual type. Three algorithms for processing such a type of joins are presented and their I/O costs are analyzed in this paper. Since such a type of joins often involves document collections of very large size, it is very important to find efficient algorithms to process them. The three algorithms differ on whether the documents themselves or the inverted files on the documents are used to process the join. Our analysis and the simulation results indicate that the relative performance of these algorithms depends on the input document collections, system characteristics, and the input query. For each algorithm, the type of input document collections with which the algorithm is likely to perform well is identified. An integrated algorithm that automatically selects the best algorithm to use is also proposed     

Optimality analysis of one-step OOSM filtering algorithms in target tracking

In centralized multisensor tracking systems, there are out-of-sequence measurements (OOSMs) frequently arising due to different time delays in communication links and varying pre-processing times at the sensor. Such OOSM arrival can induce the “negative-time measurement update” problem, which is quite common in real multisensor tracking systems. The A1 optimal update algorithm with OOSM is presented by Bar-Shalom for one-step case. However, this paper proves that the optimality of A1 algorithm is lost in direct discrete-time model (DDM) of the process noise, it holds true only in discretized continuous-time model (DCM). One better OOSM filtering algorithm for DDM case is presented. Also, another new optimal OOSM filtering algorithm, which is independent of the discrete time model of the process noise, is presented here. The performance of the two new algorithms is compared with that of A1 algorithm by Monte Carlo simulations. The effectiveness and correctness of the two proposed algorithms are validated by analysis and simulation results.     

Performance analysis of distributed deadlock detection algorithms

The paper presents a probabilistic performance analysis of a deadlock detection algorithm in distributed systems. Although there has been extensive study on deadlock detection algorithms in distributed systems, little attention has been paid to the study of the performance of these algorithms. Most work on performance study has been achieved through simulation but not through an analytic model. Min (1990), to the best of our knowledge, made the sole attempt to evaluate the performance of distributed deadlock detection algorithms analytically. Being different from Min's, our analytic approach takes the time-dependent behavior of each process into consideration rather than simply taking the mean-value estimation. Furthermore, the relation among the times when deadlocked processes become blocked is studied, which enhances the accuracy of the analysis. We measure performance metrics such as duration of deadlock, the number of algorithm invocations, and the mean waiting time of a blocked process. It is shown that the analytic estimates are nearly consistent with simulation results     

Benchmarking template-based tracking algorithms

For natural interaction with augmented reality (AR) applications, good tracking technology is key. But unlike dense stereo, optical flow or multi-view stereo, template-based tracking which is most commonly used for AR applications lacks benchmark datasets allowing a fair comparison between state-of-the-art algorithms. Until now, in order to evaluate objectively and quantitatively the performance and the robustness of template-based tracking algorithms, mainly synthetically generated image sequences were used. The evaluation is therefore often intrinsically biased. In this paper, we describe the process we carried out to perform the acquisition of real-scene image sequences with very precise and accurate ground truth poses using an industrial camera rigidly mounted on the end effector of a high-precision robotic measurement arm. For the acquisition, we considered most of the critical parameters that influence the tracking results such as: the texture richness and the texture repeatability of the objects to be tracked, the camera motion and speed, and the changes of the object scale in the images and variations of the lighting conditions over time. We designed an evaluation scheme for object detection and interframe tracking algorithms suited for AR and other computer vision applications and used the image sequences to apply this scheme to several state-of-the-art algorithms. The image sequences are freely available for testing, submitting and evaluating new template-based tracking algorithms, i.e. algorithms that detect or track a planar object in an image sequence given only one image of the object (called the template).     

AMF-MSPF: A retrospective analysis with online object tracking algorithms

Surveillance systems employ object tracking methods, a class of automatic data analysis algorithms, to accurately locate the target or its trajectory under a wide range of constraints in an uncontrolled environment. The best object tracking algorithm should be able to track under a range of constraints over long periods of time. Moreover, since object tracking algorithms are preliminary stages to these higher-level applications, they must be assigned a fraction of system resources to save computational power for more intensive higher-level decision-making tasks. Object tracking algorithms depend on a pipeline of components such as preprocessing stage, object detection, feature extraction/selection and fusion, and the search mechanism for object tracking. These multidimensional components are the tuning parameters for providing a balance between robustness, accuracy, and real-time performance of surveillance systems. In this review paper, the focus is on the detection-free class of tracking, where unlike the track-by-detection and end-to-end tracking, the object is initialized manually.The first aim of this review study is: a literature review on the multi-dimensional research trends with a focus on improving online object-tracking algorithms, and the second part is an experimental evaluation of our previously proposed Adaptive Multi-Feature Framework for MSPF (AMF-MSPF). AMF-MSPF is an MSPF-based adaptive framework that works under an online feature selection mechanism. However, a limited sequence of videos was used from the CAVIAR and PETS datasets to evaluate the AMF-MSPF. Therefore, we felt that an extensive evaluation and comparison, with online state-of-the-art tracking methods in the detection-free category, are necessary on a richer dataset to thoroughly investigate the effectiveness of AMF-MSPF and highlight its pitfalls. Twenty sequences were selected from the Amsterdam Library of Ordinary Video (ALOV++), a popular dataset with the tracking community, were selected to compare the results of the AMF-MSPF with reference methods in the online category of tracking. ALOV++ is a publically available dataset that is popular for its broad constraints, and more importantly the results of the reference methods are also publically available on this dataset. While the extensive experimentation has demonstrated the effectiveness of the AMF-MSPF against many challenging scenarios, nevertheless, this experimental analysis has highlighted some performance measures where the AMF-MSPF framework fell short of achieving even moderately satisfactory results. This research paper points toward these observations to caution researchers and advise them on what future research should concentrate on while developing generalized real-time tracking systems that would be effective under many scenarios. This research would benefit researchers and engineers working on surveillance systems to identify and tune these parameters for performance gains.     

视觉跟踪算法综述

随着信息技术与智能科学的迅速发展,计算机视觉已经成为IT产业和高新技术领域的前沿.视觉跟踪是当前计算机视觉领域的热点问题之一.阐述了视觉跟踪算法的研究现状,包括视觉跟踪算法的种类,常用数学方法,研究了基于区域的跟踪算法、基于模型的跟踪算法、基于特征的跟踪算法、基于主动轮廓的跟踪算法、参数估计方法和无参密度估计方法,并探讨了视觉跟踪算法的未来研究方向.     

Performance analysis of parallel object-oriented query processing algorithms

Advanced application domains such as computer-aided design, computer-aided software engineering, and office automation are characterized by their need to store, retrieve, and manage large quantities of data having complex structures. A number of object-oriented database management systems (OODBMS) are currently available that can effectively capture and process the complex data. The existing implementations of OODBMS outperform relational systems by maintaining and querying cross-references among related objects. However, the existing OODBMS still do not meet the efficiency requirements of advanced applications that require the execution of complex queries involving the retrieval of a large number of data objects and relationships among them. Parallel execution can significantly improve the performance of complex OO queries. In this paper, we analyze the performance of parallel OO query processing algorithms for various benchmark application domains. The application domains are characterized by specific mixes of queries of different semantic complexities. The performance of the application domains has been analyzed for various system and data parameters by running parallel programs on a 32-node transputer based parallel machine developed at the IBM Research Center at Yorktown Heights. The parallel processing algorithms, data routing techniques, and query management and control strategies have been implemented to obtain accurate estimation of controlling and processing overheads. However, generation of large complex databases for the study was impractical. Hence, the data used in the simulation have been parameterized. The parallel OO query processing algorithms analyzed in this study are based on a query graph approach rather than the traditional query tree approach. Using the query graph approach, a query is processed by simultaneously initiating the execution at several object classes, thereby, improving the parallelism. During processing, the algorithms avoid the execution of time-consuming join operations by making use of the object references among the objects. Further, the algorithms do not generate any temporary data, thereby, reducing disk accesses. This is accomplished by marking the selected objects and by employing a two-phase query processing strategy.     

分布式进化算法的性能测试与分析

针对分布式进化算法设计过程中由于缺乏对性能影响因素的分析而导致算法无法达到预期加速比的问题,提出一种全面的性能分析方法。根据分布式进化算法的组成结构,将影响分布式进化算法性能的因素分为进化操作开销、适应值计算开销和通信开销三个部分。首先研究进化算法在不同个体编码维数下进化操作开销的特性;其次,在进化操作开销相对固定的情况下,通过使用操作系统的延时函数控制适应值计算开销,通过改变个体编码维数控制通信开销;最后,应用控制变量方法,逐一测试各因素对算法加速比的影响。实验结果展现了三种因素的相互制约关系,给出了分布式进化算法获得更好加速比的条件。     

Performance analysis of multi-innovation stochastic Newton recursive algorithms

The stochastic Newton recursive algorithm is studied for system identification. The main advantage of this algorithm is that it has extensive form and may embrace more performance with flexible parameters. The primary problem is that the sample covariance matrix may be singular with numbers of model parameters and (or) no general input signal; such a situation hinders the identification process. Thus, the main contribution is adopting multi-innovation to correct the parameter estimation. This simple approach has been proven to solve the problem effectively and improve the identification accuracy. Combined with multi-innovation theory, two improved stochastic Newton recursive algorithms are then proposed for time-invariant and time-varying systems. The expressions of the parameter estimation error bounds have been derived via convergence analysis. The consistence and bounded convergence conclusions of the corresponding algorithms are drawn in detail, and the effect from innovation length and forgetting factor on the convergence property has been explained. The final illustrative examples demonstrate the effectiveness and the convergence properties of the recursive algorithms.     

Performance analysis of two different algorithms for Ethernet-FDDIinterconnection

Fiber Distributed Data Interface (FDDI) local area networks (LAN's) are used either as high-speed links between computers and peripherals, or as backbones for lower-speed LAN's, such as Ethernet and Token Ring. The availability of such a high-speed channel will lead to the implementation of high-performance distributed environments spread over a wider area than that allowed by commonly used LAN's. The performance of such distributed environments will strongly depend on that of the interconnecting devices. In this paper, two different algorithms for packet filtering are discussed, referring to bridges interconnecting Ethernet LAN's to FDDI backbones. Algorithm performance is compared with respect to 1) the traffic increase produced on a local Ethernet, and 2) the maximum allowed traffic on remote Ethernets     

Performance analysis of synchronized iterative algorithms onmultiprocessor systems

A statistical model of parallel processing and a performance evaluation technique are introduced. A task is characterized by the number of atoms and by activity. An atom is the smallest part of computation that cannot be distributed to multiple processors and all atoms of a task are assumed to be equal in computational effort. Furthermore, atoms of the task became active with a fixed probability a called the activity. The task is equally divided among processors and the computation is synchronized at periodic instances when the results can be shared. The amount of computational activity of a processor within the period between synchronizations is assumed to be a binomial random variable. The performance of the multiprocessor system is derived from the maximum order-statistic of these random variables. The theoretical performance predicted by the analysis agrees well with the reported experimental performance of logic simulation of production VLSI chips, and several observed phenomena are explainable     

类孪生网络目标跟踪算法综述

目前,在视觉目标跟踪任务中的主流方法是基于模版匹配的跟踪器,这些方法在目标的分类和边界框的回归上具有很强的鲁棒性,主要可以分为判别相关滤波跟踪器和孪生网络跟踪器,这两种方法都有一个类孪生网络的框架。以孪生网络跟踪器为例,该方法通过模版和搜索区域之间的相关操作确定目标的位置,取得了顶尖的性能表现。近年来,Transformer在计算机视觉领域的发展十分迅速,结合了Transformer的类孪生网络跟踪器在速度和精度方面都远超传统的跟踪方法。文章简要概括了判别相关滤波跟踪器、孪生网络跟踪器的发展,以及Transformer在目标跟踪任务中的应用。     

目标跟踪算法的并行优化

目标跟踪是计算机视觉领域一个重要的研究方向,近年来学者提出了众多优秀的目标跟踪算法,但许多算法的低实时性制约了其在应用场景中的有效性。针对这些算法,提出了一个通用的跟踪模型,并针对此模型提出了一个可行的并行优化方案。之后使用SCM算法验证了所提出的并行优化方案。在四核CPU的环境下,并行后的SCM算法相比于未并行的算法取得了3.48倍的并行加速比,并且比原算法Matlab+C程序的运行速度快了约30倍,这说明了所提出的并行优化方案的有效性。     

Performance analysis of parallel query processing algorithms forobject-oriented databases

Two types of parallel processing and optimization algorithms for processing object-oriented databases are the hybrid-hash pointer-based (HHP) algorithms and multi-wavefront (MWF) algorithms. We analyze these two algorithms and develop analytical formulas to capture their main performance features. We study their performance in three application environments, characterized by large databases having many object classes, each of which, respectively, (1) contains a large number of instances; (2) contains a relatively small number of instances; and (3) is of varying size. A horizontal data partitioning strategy is used in (1). A class-per-node assignment strategy is used in (2). In (3), object classes are partitioned horizontally and assigned to a varying number of processors depending on their different sizes. The MWF algorithm has three distinguishing features which contribute to its better performance: (a) a two-phase processing strategy, (b) vertical partitioning of horizontal segments, and (c) dynamic determination of the collision point in MWF propagations, which results in an optimized query execution plan. If these features are adopted by an HHP algorithm, its performance is comparable with that of the MWF algorithm because the difference in CPU time between them is negligible. The computing environment is a network of workstations having a shared-nothing architecture. The schema and some queries selected from the OO7 benchmark are used in the performance analyses and comparisons. The queries are modified slightly in different data environments in order to reflect the features of diverse database applications     

Performance analysis of locking and optimistic concurrency control algorithms

New analytic models are presented which predict the maximum throughput of locking and optimistic concurrency control algorithms for a centralized database system. By making several simplifying assumptions, these models can be easily solved. The analytic results are tested against simulation and are shown to have an accuracy considerably better than some previously reported methods. The models are used to carry out a comparison between locking and optimistic control under stated assumptions. It is found that locking schemes consistently have higher maximum throughput than optimistic schemes.     

Performance analysis of stochastic gradient algorithms under weak conditions

By using the stochastic martingale theory, convergence properties of stochastic gradient （SG） identification algorithms are studied under weak conditions. The analysis indicates that the parameter estimates by the SG algorithms consistently converge to the true parameters, as long as the information vector is persistently exciting （i.e., the data product moment matrix has a bounded condition number） and that the process noises are zero mean and uncorrelated. These results remove the strict assumptions, made in existing references, that the noise variances and high-order moments exist, and the processes are stationary and ergodic and the strong persis- tent excitation condition holds. This contribution greatly relaxes the convergence conditions of stochastic gradient algorithms. The simulation results with bounded and unbounded noise variances confirm the convergence conclusions proposed.