Particle Filter Framework Research Articles

Social interaction with visitors: mobile guide robots capable of offering a museum tour

The purpose of this study is to develop a mobile museum guide (MG) robot capable of creating and controlling spatial formations with visitors in different situations. Although much research has already been conducted in the area of nonverbal communication between guide robots and humans, the creation and controlling of spatial formations with multiple visitors is a fundamental function for MG robots that remains unexplored. Drawing upon psychological and sociological studies on the spatial relationships between humans, it is evident that to be effective MG robots should also possess the capability to create and control spatial formations in various situations. A MG robot needs to establish a spatial formation to initiate interaction with the visitors; a spatial formation is a prerequisite before the robot can begin explaining an exhibit. Moreover, the guide robot must be able to identify interested bystanders and invite them into an ongoing explanation session, necessitating a reconfiguring of the spatial formation. Finally, the robot must be able to do this while continuing to explain multiple exhibits in a cohesive fashion. To devise a system capable of meeting these needs, we began by observing and videotaping scenes of actual museum galleries. Based on analyzing these data, we found that MG creates spatial formation with the visitors in a systematic way. We then developed a mobile robot system able to create and control spatial formations while guiding multiple visitors. A particle filter framework is employed to track the visitors' positions and body orientations and the orientations of their heads. We then evaluated the guide robot system in a series of experiments that focused on different situations where a guide robot creates a spatial formation with visitors. © 2019 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

Object tracking based on learning collaborative representation with adaptive weight

A target can be lost during the tracking because of occlusion, illumination, motion, background clutter and so on. Dictionary learning methods have recently been developed to reduce the chances of missing the target; however, some dictionary learning method cannot reduce the impact between templates with high similarity very well. We developed a new approach using learning a collaborative representation with adaptive weight in a particle filter framework. The representation with collaborative dictionary can help to reduce the impact between templates with high similarity, and we employ a simple template updating strategy to get better tracking results. The experiment results on challenging sequences show that the proposed method outperforms seven state-of-the-art algorithms in terms of the overlap rate, center error and accuracy.

Mask Sparse Representation Based on Semantic Features for Thermal Infrared Target Tracking

Thermal infrared (TIR) target tracking is a challenging task as it entails learning an effective model to identify the target in the situation of poor target visibility and clutter background. The sparse representation, as a typical appearance modeling approach, has been successfully exploited in the TIR target tracking. However, the discriminative information of the target and its surrounding background is usually neglected in the sparse coding process. To address this issue, we propose a mask sparse representation (MaskSR) model, which combines sparse coding together with high-level semantic features for TIR target tracking. We first obtain the pixel-wise labeling results of the target and its surrounding background in the last frame, and then use such results to train target-specific deep networks using a supervised manner. According to the output features of the deep networks, the high-level pixel-wise discriminative map of the target area is obtained. We introduce the binarized discriminative map as a mask template to the sparse representation and develop a novel algorithm to collaboratively represent the reliable target part and unreliable target part partitioned with the mask template, which explicitly indicates different discriminant capabilities by label 1 and 0. The proposed MaskSR model controls the superiority of the reliable target part in the reconstruction process via a weighted scheme. We solve this multi-parameter constrained problem by a customized alternating direction method of multipliers (ADMM) method. This model is applied to achieve TIR target tracking in the particle filter framework. To improve the sampling effectiveness and decrease the computation cost at the same time, a discriminative particle selection strategy based on kernelized correlation filter is proposed to replace the previous random sampling for searching useful candidates. Our proposed tracking method was tested on the VOT-TIR2016 benchmark. The experiment results show that the proposed method has a significant superiority compared with various state-of-the-art methods in TIR target tracking.

Data assimilation for models with parametric uncertainty

Modeling the behavior of complex systems is notoriously difficult given approximate simulation models, parametric uncertainty, and limited and noisy data. To address such difficulty and harness information from both model forecast and observation, we propose a novel particle filter framework with the generalized polynomial chaos (gPC) method. By constructing a gPC expansion for the system state of interest, our framework delivers a system assimilation procedure that updates gPC coefficients when observations of the system are available, and whose forward model is defined by the stochastic Galerkin method. In this way, one can not only estimate the system state for specific realizations but also its statistical moments, and even the probability density function. The effectiveness of the proposed scheme is demonstrated through four numerical examples.

A Moving Target Tracking Method Based on Particle Filter and Mean-shift

Due to the limitation of using a single algorithm Mean-shift or particle filter, the deterministic tracker Mean-shift is introduced into the framework of particle filter, and a new hybrid tracking method based on particle filter and Mean-shift is proposed. The tracking method first uses Dynamic K-Means Soft Clustering algorithm to cluster multiple hypotheses, and then carries out Mean-shift deterministic search for the centers of clusters. The tracking experiments show that the proposed method can perform well when the target changes, rotates and scales, and achieves a satisfactory tracking speed.

Traffic Estimation for Large Urban Road Network with High Missing Data Ratio.

Intelligent transportation systems require the knowledge of current and forecasted traffic states for effective control of road networks. The actual traffic state has to be estimated as the existing sensors does not capture the needed state. Sensor measurements often contain missing or incomplete data as a result of communication issues, faulty sensors or cost leading to incomplete monitoring of the entire road network. This missing data poses challenges to traffic estimation approaches. In this work, a robust spatio-temporal traffic imputation approach capable of withstanding high missing data rate is presented. A particle based approach with Kriging interpolation is proposed. The performance of the particle based Kriging interpolation for different missing data ratios was investigated for a large road network comprising 1000 segments. Results indicate that the effect of missing data in a large road network can be mitigated by the Kriging interpolation within the particle filter framework.

Structured group local sparse tracker

Sparse representation is considered as a viable solution to visual tracking. In this study, the authors proposed a structured group local sparse tracker (SGLST), which exploits the local patches inside target candidates in the particle filter framework. Unlike the conventional local sparse trackers, the proposed optimisation model in SGLST not only adopts local and spatial information of the target candidates but also attains the spatial layout structure among them by employing a group-sparsity regularisation term. To solve the optimisation model, the authors proposed an efficient numerical algorithm consisting of two subproblems with the closed-form solutions. Both qualitative and quantitative evaluations on the benchmarks of challenging image sequences demonstrate the superior performance of the proposed tracker against several state-of-the-art trackers.

On-line prognosis of fatigue cracking via a regularized particle filter and guided wave monitoring

Abstract Simultaneous presence of multiple fatigue cracks in the same structural element is a critical problem affecting system safety. However, initiation and growth of these cracks is affected by various uncertainties from sources like material properties, machining quality and load conditions. Moreover, interactions among simultaneous cracks may introduce additional uncertainties, making the prognosis problem of multiple cracks more challenging. To deal with these uncertainties, this paper proposes an on-line prognostic method for multiple cracks taking advantages of the regularized particle filter and guided wave based structural health monitoring (SHM). A novel dynamic state equation of multiple cracks is proposed on the basis of the NASGRO model, which is an advanced model governing the crack growth rate from the threshold region up to the unstable crack growth region. In this state equation, a monotonic coefficient with exponentially distributed increments is proposed for each crack to determine if it starts to grow. The guided wave based SHM technique is employed to on-line collect measurements of actual crack states, which are sequentially input into the regularized particle filter framework to evaluate posterior estimations of crack states and model parameters. Based on these estimations, crack growth prognosis of each crack is performed. The fatigue test of the structure with hole-edge cracks is carried out to validate the proposed method. The result shows the effectiveness of the proposed method for identifying crack initiation and performing long-term prognosis of multiple cracks.

A data‐driven particle filter for terrain based navigation of sensor‐limited autonomous underwater vehicles

AbstractIn this article a new Data‐Driven formulation of the Particle Filter framework is proposed. The new formulation is able to learn an approximate proposal distribution from previous data. By doing so, the need to explicitly model all the disturbances that might affect the system is relaxed. Such characteristics are particularly suited for Terrain Based Navigation for sensor‐limited AUVs, where typical scenarios often include non‐negligible sources of noise affecting the system, which are unknown and hard to model. Numerical results are presented that demonstrate the superior accuracy, robustness and efficiency of the proposed Data‐Driven approach.

Robust coding in a global subspace model and its collaboration with a local model for visual tracking

The sparse representation-based trackers has attracted much attention in the research community due to its superior performance in spite of its computational complexity. But the assumption that the coding residual follows either the Gaussian or the Laplacian distribution may not accurately describe the coding residual in practical visual tracking scenarios. To deal with such issues as well as to improve the performance of the visual tracking, a novel generative tracker is proposed in a Bayesian inference framework by introducing robust coding (RC) into the PCA reconstruction. Also, it is proposed to collaborate the global and local PCA subspace appearance models to enhance the tracking performance. Further, a robust RC distance is proposed to differentiate the candidate samples from the subspace, and a novel observation likelihood is defined based on both global and local RC distances. In addition, a robust occlusion map generation and a novel appearance model update mechanism are proposed. The quantitative and qualitative performance evaluations on the OTB-50 and VOT2016 dataset demonstrate that the proposed method performs favorably against several methods based on particle filter framework.

Robust Target Tracking Algorithm Based on Superpixel Visual Attention Mechanism

As existing target tracking algorithms are prone to drift under complex environments, the authors propose a tracking algorithm based on superpixel segmentation and visual attention mechanism. The algorithm works on a particle filter framework, by conducting a superpixel segmentation first and then building a model of the visual attention mechanism for saliently mapping. After extracting HOG features of salient regions, the authors compared the HOG feature with the target template at last, in order to locate the target region in the new frame. The proposed algorithm is evaluated on a comprehensive test platform, in which the simulation show that our tracker is more efficacious and more efficient than the previous traditional target tracking algorithms to cope with target drift issue.

Object Tracking Algorithm Based on Adaptive Deep Sparse Neural Network

Due to the complexity of object tracking easy to produce the tracking drift problem, this paper proposes an object tracking algorithm based on deep sparse neural network. In the particle filter framework, using the Rectifier Linear Unit (ReLU) activation function, according to different situations of object to construct a deep sparse neural network structure, through the finite sample label on-line training, this algorithm can get a robust tracking network. The experimental results show that compared with the current mainstream tracking algorithm, the average tracking success rate and accuracy of algorithm are greatly improved, and according to changes in light, occlusion and fast object movement in complex environment, the algorithm can effectively solve the problem of tracking drift, and show good robustness.

Multi-vehicle tracking with microscopic traffic flow model-based particle filtering

This paper addresses the problem of tracking multiple vehicles on multi-lane roads with consideration for interactions among vehicles. Due to limited lane resources and traffic heterogeneity, vehicles have to interact with neighboring vehicles while moving along roads to improve their navigability and safety, resulting in highly dependent motions. However, multitarget tracking algorithms generally assume that targets move independently of one another. To address this limitation, using the microscopic traffic flow (MTF) model for modeling vehicle dynamics in the presence of interactions with surrounding traffic, an MTF-based tracking algorithm is proposed under the particle filter (PF) framework. The recursive maximum likelihood (RML) method is integrated into the PF to estimate unknown parameters in the MTF model. The posterior Cramer–Rao lower bound (PCRLB) is also derived for this problem. The performance of the proposed MTF-PF algorithm is compared with those of existing algorithms for multi-vehicle tracking on multi-lane roads. Numerical results show that the proposed algorithm requires less prior information while yielding more accurate and consistent tracks.

Multi-task non-negative matrix factorization for visual object tracking

This paper proposes an online object tracking algorithm in which the object tracking is achieved by using multi-task sparse learning and non-negative matrix factorization under the particle filtering framework. The object appearance is first modeled by subspace learning to reflect the target variations across frames. Combination of non-negative components is learned from examples observed in previous frames. In order to robust tracking an object, group sparsity constraints are included to the non-negativity one. Furthermore, the alternating direction method of multipliers algorithm is employed to compute the model efficiently. Qualitative and quantitative experiments on a variety of challenging sequences show favorable performance of the proposed algorithm against state-of-the-art methods.

Particle filtering of dynamical networks: Highlighting observability issues.

In a network of high-dimensionality, it is not feasible to measure every single node. Thus, an important goal is to define the optimal choice of sensor nodes that provides a reliable state reconstruction of the network system state-space. This is an observability problem. In this paper, we propose a particle filtering (PF) framework as a way to assess observability properties of a dynamical network, where each node is composed of an individual dynamical system. The PF framework is applied to two benchmarks, networks of Kuramoto and Rössler oscillators, to investigate how the interplay between dynamics and topology impacts the network observability. Based on the numerical results, we conjecture that, when the network nodal dynamics are heterogeneous, better observability is conveyed for sets of sensor nodes that share some dynamical affinity to its neighbourhood. Moreover, we also investigate how the choice of an internal measured variable of a multidimensional sensor node affects the PF performance. The PF framework effectiveness as an observability measure is compared with a well-consolidated nonlinear observability metric for a small network case and some chaotic system benchmarks.

Object tracking via dual fuzzy low-rank approximation

For the object loss problem in the tracking process caused by illumination, occlusion, pose variation, and motion blur, the tracking method based on dual fuzzy low-rank approximation in a particle filter framework is proposed in this paper. Firstly, multiple constraint regions are built to filter insignificant samples, and more distinguished candidate samples are selected. Secondly, dual fuzzy observation function of each candidate sample is created based on the designed low-rank approximation representations of object and background. Then the generalized tracking results are obtained by computing membership degrees of dual fuzzy observation functions. Finally, based on the spatial coherency principle, the final tracking result is determined from the generalized results by measuring similarities of consecutive objects. The proposed method shows good performance as compared with several state-of-the-art trackers on challenging benchmark sequences.

Visual object tracking via a manifold regularized discriminative dual dictionary model

Real-time object tracking plays an important role in many computer vision systems, yet in complex scenarios, it is still a very challenging problem. In this paper, we propose a new visual tracking algorithm via a manifold regularized discriminative dual dictionary (MD3) model. First, a dual dictionary is introduced to avoid the calculation of representation coefficient in distance function construction. Second, the local background templates are utilized to keep the learned dictionaries discriminative. Third, the manifold regularization on representation coefficient is proposed to ensure that MD3 model has a bit error tolerance on the object update. We formulate object tracking in a particle filter framework, in which the observation model is calculated as the reconstruction error between learned dictionaries and the candidate template. Extensive experiments in various tracking scenarios are performed to evaluate the proposed method, and the results interpret that the tracking accuracy as well as the computational cost can be improved as compared with the state-of-the-art approaches.

Neuro-probabilistic model for object tracking

Occlusion is one of the major challenges for object tracking in real-life scenario. Various techniques in particle filter framework have been developed to solve this problem. This framework depends on two issues, namely motion model and observation (i.e., likelihood) model. Due to the lack of effective observation model and efficient motion model, problem of occlusion still remains unsolvable in the tracking task. In this article, an effective observation model is proposed based on confidence (classification) score provided by the developing online probabilistic neural network-based discriminative appearance model. The appearance model is trained with the prior knowledge of two classes (i.e., object and background) and tries to discriminate between three classes, namely object, background and occluded part of the object. The considered composite motion model can handle both the object motion and scale change in the object. The proposed update mechanism is able to adapt the appearance change in an object during tracking. We show a realization of the proposed method and demonstrate its performance (both quantitatively and qualitatively) with respect to state-of-the-art techniques on several challenging sequences. Analysis of the results concludes that the proposed technique can track fully (or partially) occluded object as well as object in various complex environments in a better way as compared to the existing ones.

Robust Object Tracking Using Manifold Regularized Convolutional Neural Networks

In visual tracking, usually only a small number of samples are labeled, and most existing deep learning based trackers ignore abundant unlabeled samples that could provide additional information for deep trackers to boost their tracking performance. An intuitive way to explain unlabeled data is to incorporate manifold regularization into the common classification loss functions, but the high computational cost may prohibit those deep trackers from practical applications. To overcome this issue, we propose a two-stage approach to a deep tracker that takes into account both labeled and unlabeled samples. The annotation of unlabeled samples is propagated from its labeled neighbors first by exploring the manifold space that these samples are assumed to lie in. Then, we refine it by training a deep convolutional neural network using both labeled and unlabeled data in a supervised manner. Online visual tracking is further carried out under the framework of particle filters with the presented manifold regularized deep model being updated every few frames. Experimental results on different tracking datasets demonstrate that our tracker outperforms most existing tracking approaches. The source code and results are available at: https://github.com/shenjianbing/MRCNNTracking .

Robust information unscented particle filter based on M‐estimate

Herein, inspired by the unscented particle filter (UPF), an improved particle filter (PF) using an unscented information filter based on M-estimate (UIF-BM) to approximate the importance density function (IDF) is designed. The filtering method is proposed to obtain an accurate state estimation for a non-linear discrete time dynamical system with non-Gaussian system noises and observation noises contaminated by some Gaussian impulsive noises. The PF framework is used as a common method to handle non-Gaussian system noises. In order to obtain a precise IDF, UIF based on M-estimate is used for heavy-tailed observation noises. Meanwhile, instead of the robust Gaussian filter, the robust IF can avoid the numerical problem that zero weight functions cannot be incorporated into the framework. The simulation results indicate the estimation accuracy and efficiency of the proposed filter. Compared with the UIF-BM, PF, and UPF, the superiority of the proposed filter against the non-ideal system and observation noises is obvious.