Robust shape tracking with multiple models in ultrasound images

This paper addresses object tracking in ultrasound images using a robust multiple model tracker. The proposed tracker has the following features: 1) it uses multiple dynamic models to track the evolution of the object boundary, and 2) it models invalid observations (outliers), reducing their influence on the shape estimates. The problem considered in this paper is the tracking of the left ventricle which is known to be a challenging problem. The heart motion presents two phases (diastole and systole) with different dynamics, the multiple models used in this tracker try to solve this difficulty. In addition, ultrasound images are corrupted by strong multiplicative noise which prevents the use of standard deformable models. Robust estimation techniques are used to address this difficulty. The multiple model data association (MMDA) tracker proposed in this paper is based on a bank of nonlinear filters, organized in a tree structure. The algorithm determines which model is active at each instant of time and updates its state by propagating the probability distribution, using robust estimation techniques.     

IMM算法到systolic阵的映射研究

ＩＭＭ算法是一种性能优良的机动目标跟踪算法。但其计算量比较厉,所以实时实现比较困难,这就限制了它的应用。但ＩＭＭ算法具有非常好的并行性适合行并行实现。而ｓｙｓｔｏｌｉｃ阵是一种典型的并行处理结构。所以,本文将ＩＭＭ算法映射到ｓｙｓｔｏｌｉｃ阵列结构上,从而实现了ＩＭＭ算法的并行处理,达到了实时实现的目的。分析了结果表明,本文提出了真有加速比大、扩展性好的特点。     

基于模板匹配的视频对象分割

视频对象分割是MPEG-4标准关键技术.本文结合模板匹配和基于运动估值和补偿的对象跟踪方法,提出了一种可以从复杂场景中分割出MPEG-4视频对象的新方法.在使用运动估值和补偿得到分割掩膜后,以初始帧对象颜色为模板,在当前帧的轮廓边界区域通过模板匹配检测对象,使轮廓精确化.本文方法在一定范围内有效解决了遮挡问题,并能够以初始帧跟踪任意长序列中的对象.     

基于奇异值分解的图像目标跟踪算法

传统相关跟踪方法是利用模板图像与目标图像对应像素的灰度差异信息进行跟踪,它对旋转变化敏感,且存在跟踪累积误差,容易导致模板漂移而丢失目标。文中提出基于奇异值分解的跟踪算法,算法首先建立模板图像训练集合,利用奇异值分解方法,张成模板图像特征空间,然后求出模板图像在特征空间里的投影值,代替传统算法中灰度对两幅待匹配图像进行的全局搜索定位。在进行投影值间的相似性度量时,欧氏距离同等对待所有的特征向量不移合理,文中采用了一种鲁棒估计方法,可以对不同距离的值做不同处理。匹配跟踪实验效果良好。     

基于模型调整的自适应交互多模型算法

针对交互多模型算法的技术特点,利用量测中所包含的当前信息对目标模型集的自适应调整并对调整后的模型概率进行估计,实现了模型集的自适应交互多模型算法。介绍了这种算法的设计步骤和仿真方法。仿真结果表明了该算法比标准IMM算法具有更高精度的跟踪性能。     

Self-tuning fault diagnosis of MEMS

Multiple-model adaptive estimation techniques have been previously successfully applied to fault diagnosis of microsystems. Their diagnosis performance highly depends on the accuracy of modeling techniques used in representing faults. This paper presents the application of a self-tuning forgetting factor technique in the modeling of faults in MEMS and its effects on diagnosis performance compared with the application of Kalman filters and fixed gain estimation techniques. The self-tuning-based modeling used in the diagnosis algorithm was experimentally implemented. It demonstrated superior results compared to Kalman filter and fixed gain estimation techniques by accelerating the diagnosis process.     

自适应转移概率交互式多模型跟踪算法

针对标准的交互式多模型算法（Interacting Multiple Model,IMM）存在模型集设计困难和采用固定转移概率矩阵导致模型切换缓慢、跟踪精度下降的不足,提出一种自适应转移概率IMM算法.首先,提出了一种新的模型集设计方法,将强跟踪修正输入估计（Strong Tracking Modified Input Estimation,STMIE）模型和匀速运动（Constant Velocity,CV）模型作为IMM算法的模型集,利用STMIE算法对高机动目标的跟踪能力以及CV模型对非机动目标跟踪的高精度,实现对目标的全面自适应跟踪.其次,提出一种依据模型似然函数值对Markov转移概率进行实时修正的方法,增强匹配模型的作用,削弱不匹配模型的影响.仿真结果表明,依据模型似然函数修正转移概率的方法使IMM算法的模型切换速度和跟踪精度都得到提高,提出的IMM-STMIECV算法的跟踪精度高于IMM-CVCA、IMM-CVCACT以及IMM-CVCS算法.     

基于IMM-UKF的机动目标跟踪方法

针对目标运动模式的不确定性和运动模型的非线性问题,提出基于无味卡尔曼滤波器的交互式多模型方法。该算法采用匀速运动模型、匀加速运动模型、已知转弯角速度的匀速率转弯模型和“当前”统计模型作为模型集,用无味卡尔曼滤波实现非线性状态估计。仿真结果表明,该算法比传统的交互式多模型算法具有更高的跟踪性能。     

Probabilistic Kalman filter for moving object tracking

Kalman filter has been successfully applied to tracking moving objects in real-time situations. However, the filter cannot take into account the existing prior knowledge to improve its predictions. In the moving object tracking, the trajectories of multiple targets in the same environment could be available, which can be viewed as the prior knowledge for the tracking procedure. This paper presents the probabilistic Kalman filter (PKF) that is able to take into account the stored trajectories to improve tracking estimation. The PKF has an extra stage after two steps of the Kalman filter to refine the estimated position of the targets. The refinement is obtained by applying the Viterbi algorithm to a probabilistic graph, that is constructed based on the observed trajectories. The graph is built in the offline situation and could be adapted in the online tracking. The proposed tracker has higher accuracy compared to the standard Kalman filter and could handle widespread problems such as occlusion. Another significant achievement of the proposed tracker is to track an object with anomalous behaviors by drawing an inference based on the constructed probabilistic graph. The PKF was applied to several manually-built videos and several other video-bases containing severe occlusions, which demonstrates a significant performance in comparison with other state-of-the-art trackers.     

红外序列图像目标跟踪的自适应Kalman滤波方法

提出了一种用于动态序列图像目标跟踪的自适应Kalman滤波方法。该方法用函数估计的思想估计目标的当前运动模型，同时实时修改滤波器的统计模型，并将最小二乘支持向量机应用于对当前目标运动模型的估计。实验表明，此种改进的Kalman滤波器的算法在跟踪机动目标时具有良好的性能。     

基于改进IMMPF的空间激光通信移动平台跟踪方法研究

基于空间移动平台的空间激光通信是实现未来超大容量空间通信的主要途径,而高精度空间平台动态跟踪问题一直是其研究的难点之一。空间移动平台跟踪是一个纯角度的非线性跟踪,用单一的模型很难准确描述空间移动平台的运动状态,需要采用多个模型描述机载平台的运动状态,但交互多模型估计(IMM)需用先验知识和复杂计算量,研究将模型转移概率与模糊推理相结合解决测量空间的不确定性和复杂计算量。标准粒子滤波(PF)能够处理非线性问题,但会引起粒子滤波器粒子退化和发散,采用无迹粒子滤波(UPF)克服了标准粒子滤波没有考虑最新量测信息的问题。仿真结果表明模糊交互多模型粒子滤波算法能较大地提高跟踪性能。     

Novel extended Viterbi-based multiple-model algorithms for state estimation of discrete-time systems with Markov jump parameters

This paper considers a state estimation problem for discrete-time systems with Markov switching parameters. For this, the generalized pseudo-Bayesian second-order-extended Viterbi (GPB2-EV) and the interacting multiple-model-extended Viterbi (IMM-EV) algorithms are presented. The derivations of these new algorithms rely on a nontrival incorporation of some functional mechanisms of a new extended Viterbi algorithm into the GPB2 and the IMM methods for hypothesis reductions in order to improve computational efficiency and/or estimation performance. The IMM-EV (and the GPB2-EV) algorithm and the IMM (and the GPB2) algorithm have some common components, but their schemes for the calculation of weights and for the combination of the inputs and outputs are different. Indeed, the IMM-EV (and the GPB2-EV) algorithm spans the continuum from hard-decision methods with merged-hypothesis-tree style to the IMM (and the GPB2) algorithm inclusive. The proposed algorithms are well suited to state estimation problems in maneuvering target tracking. Simulations demonstrate that an IMM-EV algorithm can be an improvement to the IMM, the GPB2, and the variable-structure multiple model with likely-model set methods for tracking a target undergoing various types of maneuvers at some unknown times.     

基于摄像机运动控制的实时对象检测与跟踪算法

提出了一种基于摄像机运动控制的实时运动对象检测与跟踪算法.该算法首先采用非参数核密度估计方法在复杂的动态背景条件下检测运动对象区域;然后由CamShift算法计算跟踪目标的位置,并采用Kalman滤波预测对象的运动信息来控制摄像机的运动,能够准确地跟踪对象并有效解决了背景中的部分遮挡问题.在SonyRZ25网络摄像机上完成了对象的实时检测与跟踪,对多个室内和室外场景的实验验证了所提出方法的实用性和有效性.     

基于SRCDKF的交互式多模型算法

给出了一种基于平方根中心差分卡尔曼滤波的交互多模型(IMM)算法.该算法较好地解决了非线性条件下机动目标跟踪的问题,可获得比基于扩展卡尔曼滤波(EKF)的IMM算法更好的数值稳定性、计算精度和收敛速度,还避免了复杂的Jacobi矩阵运算.Monte Carlo仿真验证了该算法的可行性和有效性.     

对海面远距离运动目标定位跟踪的IMM算法研究

研究了平流层三站测时差无源定位系统对海面远距离运动目标定位与跟踪的滤波算法问题。将IMM滤波器应用到平流层三站测时差定位系统中,采用三模型组合的IMM滤波算法,首先通过时差测量参数计算出粗略的定位结果,接下来采用IMM滤波算法逐步估计出运动辐射源的速度,并修正初始的定位结果,提高定位精度。仿真结果表明,基于IMM滤波器的三站时差定位系统具有较快的收敛时间、较高的跟踪精度和适中的计算复杂度,能够对不同强度的机动目标进行定位与跟踪,很好地解决了三站时差定位系统对海面远距离运动辐射源定位跟踪的算法问题。还提出了利用回归处理的方法进一步提高固定目标定位精度的思想。     

基于运动学参数的高速高机动目标跟踪自适应IMM算法

提出一种基于运动学参数的自适应IMM算法,进行高速高机动目标跟踪.该算法利用交互输出的结果,计算目标的运动学参数,进而实时估算出机动目标的转弯率,利用估计值自适应地使模型与目标实际运动状态匹配.文中还将该算法与其他两种IMM算法进行了比较.蒙特卡罗仿真结果表明,这种算法跟踪目标的稳定性和精确性均优于其他两种算法,有利于高速高机动目标的实时跟踪,具有较好的可实现性.     

Human pose estimation and its application to action recognition: A survey

Human pose estimation aims at predicting the poses of human body parts in images or videos. Since pose motions are often driven by some specific human actions, knowing the body pose of a human is critical for action recognition. This survey focuses on recent progress of human pose estimation and its application to action recognition. We attempt to provide a comprehensive review of recent bottom-up and top-down deep human pose estimation models, as well as how pose estimation systems can be used for action recognition. Thanks to the availability of commodity depth sensors like Kinect and its capability for skeletal tracking, there has been a large body of literature on 3D skeleton-based action recognition, and there are already survey papers such as [1] about this topic. In this survey, we focus on 2D skeleton-based action recognition where the human poses are estimated from regular RGB images instead of depth images. We summarize the performance of recent action recognition methods that use pose estimated from color images as input, then show that there is much room for improvements in this direction.     

使用交互多模型算法的高速高机动目标跟踪

对机动目标进行跟踪一直是甚具挑战性的问题,特别是跟踪高速高机动目标在理论上和实践上都有较高的技术难度。现有各种算法在这个问题上均有各自的缺点和不足。该文在现有的运动机动模型和IMM算法的基础上,提出了使用多种机动模型交互的IMM算法进行高速高机动目标跟踪。不同机动模型之间的互补使这种算法克服了使用单一模型的一些问题。使用“当前”统计模型、二级滤波模型和CV模型进行交互是一种可行的高速高机动目标跟踪方案。为验证算法的有效性,进行了Monte Carlo仿真。仿真结果表明,该算法在性能和计算复杂度之间取得了较好的平衡,有很好的可实现性。     

A New Multiple Model Filter With Switch Time Conditions

The interacting multiple model filter has long been the method of choice for performing target tracking using multiple motion models. The filter finds a suboptimal solution to a problem which has the implicit assumption that immediate model shifts have the highest probability. When the sampling rate of the underlying continuous process is high compared to the target dynamics, this is not a reasonable assumption. Instead, changes in dynamics persist for some time. In this paper we propose an alternative switching model, which forces the dynamic models to persist for at least a model-specific time. The model is semi-Markov in nature, with a sojourn time probability mass function which is zero for a model-specific number of time steps, and then follows a geometrical distribution. Through this assumption a less complex problem in terms of model hypotheses arises, and to that problem we derive a state estimation algorithm that is close to optimal when the model assumptions are valid. Three other semi-Markov-based multiple-model filters are discussed and compared to in a qualitative sense. We also derive a new aircraft motion model for start and termination of turns. Finally, the proposed filter is evaluated on a benchmark scenario for tracking, and the results show a performance increase compared to the interacting multiple model (IMM) filter for the trajectories considered.     

Adaptive Reibungskompensation zur Verbesserung des Folgeverhaltens bei Visual Servoing

The vollowing article surveys the application of adaptive friction compensation to improve tracking behaviour using classical methods of motion control for visual servoing. Visual servoing is the use of motion control in order to follow a moving object, e.g. by the use of a robot and a gripping device. The position of the object is captured through image processing. The survey has been carried out on an XY-linear axis system using simulations and real time measurements. It shows that the tracking behaviour with classical motion control using visual servoing can be considerably improved by adaptive friction compensation. Problems which occur using such methods of motion control with visual servoing are also displayed.