Benlian Xu

Ant clustering PHD filter for multiple-target tracking

A novel ant clustering filtering algorithm, under the guidance of first-order statistic moment of posterior multiple-target state (probability hypothesis density), is investigated and applied to estimate the time-varying number of targets and their individual states in a cluttered environment. The ant clustering filtering algorithm includes two clustering steps: the first step is called rough ant clustering, which involves the stochastic selection of each ant and its state local adjustment according to the current likelihood function and posterior intensity, respectively; while the second is called fine ant clustering, which employs these ants to extract the multiple-target state. Numerical simulations verify the tracking multiple-target capability of our proposed algorithm through performance comparison with the Sequential Monte Carlo (SMC) method.

A novel estimator with moving ants

Abstract In the traditional ant colony optimization (ACO) algorithm, ants are utilized to solve various combinatorial optimization problems regardless of their individual velocities. In this work, however, a moving ant estimator (MAE) is developed on the premise that the issue of moving velocity of each ant is considered. The determination of the velocity of each ant depends directly on the normalized weights between the one-step prediction of the ant starting position and the selected one-step position. Besides this, the velocity of each ant is further adjusted locally based on its individual moving ability, which is defined in the corresponding “pheromone” update process just as the traditional ACO algorithm. To improve the accuracy of the proposed estimator, two improved versions are investigated. Numerical simulations show that the moving ant estimator, as well as its two improved versions, could estimate adaptively the state of maneuvering or non-maneuvering target. To verify the effectiveness of the MAE, the obtained results are compared with those from PF, IMMPF, etc.

An accurate multi-cell parameter estimate algorithm with heuristically restrictive ant system

To reliably analyze multi-cell motion in a series of low-contrast image sequences, we present a novel heuristically restrictive ant system, which operates in a non-optimization way, to adaptively estimate multiple parameters of multiple cells. First, the local intensity variation measure on each pixel of image is defined to generate ant colony initial distribution positions, which are further treated as boundary markers to restrict ant searching behavior. Afterwards, to speed up the ant searching process, both location and contour ant decision behaviors are modeled appropriately to acquire cell position and edge estimates on their individual pheromone fields, which are formed by restrictive pheromone deposits but operate independently and in parallel. Finally, the stability of our proposed pheromone control mechanism is proven to guarantee reliable multi-parameter extraction. Experiment results show that our algorithm could automatically and accurately track numerous cells in various scenarios, and it shows considerable robustness against other popular tracking methods. Our algorithm can track multiple cells with the variance in morphology, the difference in dynamics and the changes in number.Our algorithm could give a joint estimate of state and contour of each cell even without birth ant colonies.Our algorithm enjoys a robust tracking performance with high percentage of tracked position when comparing with existing methods.

An ant stochastic decision based particle filter and its convergence

Particle filter (PF) is a kind of flexible and powerful sequential Monte-Carlo technique designed to solve the optimal nonlinear parameter estimation numerically, and the degradation of particles in generic PF occurs when it is applied to the model switching dynamic system. To avoid this phenomenon, an ant stochastic decision based particle filter is proposed to encapsulate model switching information through dividing probabilistically particles into two model operations, and then a well defined re-sampling scheme is introduced to gain a better overlap with the true density function. To show the theoretic consistency with the generic PF, its basic convergence result is presented as well. Finally, we compare the performance of our proposed algorithm with that of other estimators (e.g., PF and moving ant estimator), and simulation results demonstrate its superior robustness of parameter estimation for switching dynamic system.

Ants for track initiation of bearings-only tracking

Abstract A novel track initiation method using ants of different tasks, a kind of ant colony optimization (ACO) algorithm, is developed in this paper. For the proposed system of ants of different tasks, we assume that the number of tracks to be initiated equals the one of tasks, and moreover, ants of the same task search for a given track by collaboration, while ants of different tasks will compete with each other during the search process. In order to fulfill such behaviors, the pheromone model is established, and the corresponding objective function to be optimized is also presented. Numerical simulation results indicate that, for the case of bearings-only multi-sensor-multi-target tracking, the track initiation performance for the proposed system of ants of different tasks performs well compared to other track initiation methods.

On the Cramér-Rao lower bound for biased bearings-only maneuvering target tracking

This paper aims to investigate the theoretical performance of bearings-only maneuvering target tracking when fixed measurement biases exist. Firstly, the general formulation of Cramer-Rao lower bound (CRLB) is presented when both targets maneuver and fixed measurement biases work simultaneously in a bistatic system, and the corresponding recursive CRLB formulation is then derived. Finally, a maneuver detection method is developed to calculate the approximation of the theoretical CRLB. Numerical simulation results indicate either that the proposed or the fuzzy-neural-network-based maneuver detection method can yield a satisfying CRLB compared with the theoretical CRLB, but the proposed method is easier to implement owing to directly utilizing the targets bearing information.

An ant-based stochastic searching behavior parameter estimate algorithm for multiple cells tracking

This paper presents a novel ant-based parameter estimate algorithm to accurately track multiple cells in a series of low-contrast image sequences. Our proposed algorithm consists of three main blocks, i.e., priori colony distribution block, multi-colony reconstruction block, and cell labeling and state extraction block. Priori colony distribution block aims to directly distribute birth ants into regions where cells probably occur, which is implemented through kernel density probability estimate. Multi-colony reconstruction block is to move ants towards potential regions based on histogram similarity and place agent pheromone with appropriate introduction to evaporation and propagation models. Cell labeling and state extraction block is implemented by a fast ant clustering algorithm to determine the number of cells and their individual states, and the ratio of known identity ants to unknown ants in a cluster contributes to discriminate cell identity. Experiment results show that our algorithm could automatically track numerous cells in various scenarios, and furthermore, it is more accurate and robust than other popular tracking methods.

Modeling analysis of ant system with multiple tasks and its application to spatially adjacent cell state estimate

The problem of multi-cell tracking plays an important role in studying dynamic cell cycle behaviors. In this paper, a novel ant system with multiple tasks is modeled for jointly estimating the number of cells and individual states in cell image sequences. In our ant system, in addition to pure cooperative mechanism used in traditional ant colony optimization algorithm, we model and investigate another two types of ant working modes, namely, dual competitive mode and interactive mode with cooperation and competition to evaluate the tracking performance on spatially adjacent cells. For adjacent ant colonies, dual competitive mode encourages ant colonies with different tasks to work independently, whereas the interactive mode introduces a trade-off between cooperation and competition. In simulations of real cell image sequences, the multi-tasking ant system integrated with interactive mode yielded better tracking results than systems adopting pure cooperation or dual competition alone, both of which cause tracking failures by under-estimating and over-estimating the number of cells, respectively. Furthermore, the results suggest that our algorithm can automatically and accurately track numerous cells in various scenarios, and is competitive with state-of-the-art multi-cell tracking methods.

A real-time moving ant estimator for bearings-only tracking

A real-time moving ant estimator (RMAE) is developed for the bearings-only target tracking, in which ants located at their individual current state utilize the normalized weight and pheromone value to select the one-step prediction state and the dynamic moving velocity of each ant is depended directly on the normalized weights between two states Besides this, two pheromone update strategy is implemented Numerical simulations indicate that the RMAE could estimate adaptively the state of maneuvering or non-maneuvering target, and real-time requirement is superior to the moving ant estimator (MAE).

Analysis and approximation of performance bound for two-observer bearings-only tracking

This paper presents the analytic recursive formulas of Cramer-Rao lower bound (CRLB) for the switching models system, in which the target moves either with a constant velocity or with a constant speed and a constant turn rate. For the case of two-observer bearings-only maneuvering target tracking, a reliable maneuver detection method is investigated and then utilized to approximate the theoretic CRLB. Finally, to demonstrate the agreement between the approximated CRLB using the proposed maneuver detection method and the theoretic one, a large number of Monte-Carlo runs under different maneuvering scenarios are conducted. Correctness of the analytic recursive formulas of CRLB and effectiveness of the proposed maneuver detection method are verified from these simulations.