We employ hierarchical data association to track players in team sports. Player movements are often complex and highly correlated with both nearby and distant players. A single model would require ...many degrees of freedom to represent the full motion diversity and could be difficult to use in practice. Instead, we introduce a set of Game Context Features extracted from noisy detections to describe the current state of the match, such as how the players are spatially distributed. Our assumption is that players react to the current situation in only a finite number of ways. As a result, we are able to select an appropriate simplified affinity model for each player and time instant using a random decision forest based on current track and game context features. Our context-conditioned motion models implicitly incorporate complex inter-object correlations while remaining tractable. We demonstrate significant performance improvements over existing multi-target tracking algorithms on basketball and field hockey sequences several minutes in duration and containing 10 and 20 players respectively.
This paper proposes a novel single vehicle tracking algorithm with enhanced reliability for automotive radar systems. The proposed algorithm overcomes the weaknesses of the probabilistic data ...association filter (PDAF) in single-target tracking in clutter. The PDAF is successful in normal situations, but may fail to track a target owing to various factors, such as the initialization errors and the sudden changes in the target motion. The proposed algorithm can recover the PDAF from failures using an assisting finite impulse response (FIR) filter. The FIR filter operates only when the PDAF cannot track a target properly, and additionally offers state estimate and estimation error covariance to reset the PDAF. The proposed algorithm, the hybrid PDAF/FIR filter (HPFF), combines the PDAF and FIR filter, and hence shows enhanced reliability. Simulations of preceding vehicle tracking using an automotive radar demonstrate the effect and performance of the proposed HPFF.
We propose a framework for tracking multiple targets, where the input is a set of candidate regions in each frame, as obtained from a state-of-the-art background segmentation module, and the goal is ...to recover trajectories of targets over time. Due to occlusions by targets and static objects, as also by noisy segmentation and false alarms, one foreground region may not correspond to one target faithfully. Therefore, the one-to-one assumption used in most data association algorithms is not always satisfied. Our method overcomes the one-to-one assumption by formulating the visual tracking problem in terms of finding the best spatial and temporal association of observations, which maximizes the consistency of both motion and appearance of trajectories. To avoid enumerating all possible solutions, we take a data-driven Markov Chain Monte Carlo (DD-MCMC) approach to sample the solution space efficiently. The sampling is driven by an informed proposal scheme controlled by a joint probability model combining motion and appearance. Comparative experiments with quantitative evaluations are provided.
A modified version of the probabilistic data association (PDA) is proposed for target tracking under measurement uncertainty conditions. This method uses the likelihood of each validated measurement, ...and selects the best k candidates to be integrated with the PDA. Different computer simulations with a single target under dense-cluttered environment are presented. Comparison to the standard PDA shows a track loss reduction with the proposed method.
Multiple particle tracking (MPT) has seen numerous applications in live-cell imaging studies of subcellular dynamics. Establishing correspondence between particles in a sequence of frames with high ...particle density, particles merging and splitting, particles entering and exiting the frame, temporary particle disappearance, and an ill-performing detection algorithm is the most challenging part of MPT. Here we propose a tracking method based on multidimensional assignment to address these problems. We combine an Interacting Multiple Model (IMM) filter, multidimensional assignment, particle occlusion handling, and merge-split event detection in a single software analysis package. The main advantage of a multidimensional assignment is that both spatial and temporal information can be used by using several later frames as reference. The IMM filter, which is used to maintain and predict the state of each track, contains several models which correspond to different types of biologically realistic movements. It works especially well with multidimensional assignment, because there tends to be a higher probability of correct particle association over time. First the method generates many particle-correspondence hypotheses, merge-split hypotheses and misdetection hypotheses within the framework of a sliding window over the frames of the image sequence. Then it builds a multidimensional assignment problem (MAP) accordingly. The particle is tracked with gap-filling, and merging and splitting events are then detected using the MAP solution. The tracking method is validated on both simulated tracks and microscopy image sequences. The results of these experiments show that the method is more accurate and robust than other “tracking from detected features” methods in dense particle situations.
Data association is significantly important for SLAM. The correct data association may improve the accuracy of localization. In this paper we present a new approach for the Data Association problem ...of SLAM, based upon K-means Clustering and JCBB. First, The measurements at every step are separated into several groups. The number of groups is decided by the characteristics of the environment. JCBB and ICNN is then used on each group for getting several local matching results. Finally, we combine the local matching results by JCBB or ICNN together and find result that has the best joint compatibility as the best global matching result. The experiment results show that this approach can achieve that the matching accuracy is similar to JCBB and the matching time is far better than JCBB.
This paper proposes a new intuitionistic fuzzy joint probabilistic data association filter for multitarget tracking in a cluttered environment. In the proposed algorithm, the joint association ...probabilities in JPDAF are reconstructed by utilizing intuitionistic fuzzy membership degrees of the measurements belonging to the targets. To compute the intuitionistic fuzzy membership degree, a new intuitionistic fuzzy clustering method is proposed based on intuitionistic fuzzy point operator, which can extract useful information from uncertainty information of measurement. At the same time, two new weight assignments are introduced to deal with the uncertainty of measurement, which lead to two different data association methods, IF-JPDAF1 and IF-JPDAF2. Moreover, according to the characteristic of multitarget tracking, a new intuitionistic index of intuitionistic fuzzy set is defined. Finally, experiment results show the proposed algorithms have advantages over the conventional methods (including the JPDAF, Fitzgerald's JPDAF and MEF-JPDAF) in terms of efficiency and robustness.
•A fuzzy clustering is proposed based on intuitionistic fuzzy (IF) point operator.•The joint association probabilities are reconstructed by utilizing IF membership.•A new intuitionistic index of intuitionistic fuzzy set is defined.•This paper proposes two different methods (IF-JPDAF1 and IF-JPDAF2).
Ultra-Wide Band (UWB) radar has a number of advantages of resolving multipath, exceptional spatial resolution, and ranging performance. However, several difficulties are confronted for multiple ...target tracking using UWB radars such as clutter signals which contaminate target signals, and unidentified number and behavior of the targets. Hence, this paper presents to develop a multiple moving target tracking algorithm, consisting of preprocessing and multiple target tracking steps. In the preprocessing step, static clutter reduction and constant false alarm rate (CFAR) detection extract the target candidate range measurements from each UWB radar. Then, two multiple target tracking (MTT) steps are developed: range- based MTT and position-based MTT. The range-based MTT is mainly based on existing linear multi-target integrated probabilistic data association (LM-IPDA) from each UWB radar measurement. Then the outputs of each LM-IPDA are gathered in the positioning center to estimate the position of multiple targets. On the other hands, the position-based MTT is based on multiple sensor LM-IPDA (msLM-IPDA) as an accurate target tracking method for various uncertainties by improving the probabilistic model of LM-IPDA. The tracking performance of two MTT methods is investigated with both numerical simulation and experiments.
Obtaining reliable and discriminative target representation are two vital tasks for data association in multi-tracking. Pervious works always directly combine bunch of features for more ...discriminative target representation, but this is prone to error accumulation and unnecessary computational cost, which on the contrary may increase identity switches in data association. Moreover, reliability of a same feature in different scenes may vary a lot, especially for currently widespread network cameras, which have been settled in complex and various scenes, previous fixed feature selection scheme cannot meet general requirements. To address this problem, we propose a scene-adaptive hierarchical data association scheme, which adaptively selects features which have higher reliability on target representation in applied scene, and gradually combines features to the minimum requirements of discriminating ambiguous targets. Hierarchical feature space is constructed according to reliability of features in the multi-tracking system, and data association is conducted in different layers of the feature space adaptively. Our algorithm is validated on various challenging RGB-D and RGB datasets recorded in various indoor and outdoor scenes, for diversities of both features and scenes. Experimental results validate its effectiveness and efficiency.
To deal with the track coalescence problem of the joint probabilistic data association (JPDA) filter, a novel approach based on the Kullback–Leibler divergence (KLD) is developed in this study. In ...JPDA, the posterior probability density function (PDF) is approximated by a single Gaussian PDF at each time step. The authors propose a novel method of optimising the posterior PDF to obtain a single Gaussian PDF that minimises the KLD from the posterior PDF. However, the KLD is intractable because the posterior PDF is a Gaussian mixture model. Hence, an approximation of the KLD is introduced as the cost function to simplify the problem. The cost function is a linear combination of multiple objective functions which are not conflicting. Therefore, the minimisation of the cost function is easier to operate, because all objective functions can be optimised simultaneously. In addition, an iterative method is adopted for minimising the cost function. In the iteration process, the tracking accuracy is improved with the monotonic decrease of the cost function. Theoretical analysis and example show the feasibility of the proposed approach. Simulation results demonstrate the advantages of the new approach over others when tracking closely spaced targets with contaminated sensor measurements.