Search

An approach to automatically detect bacteria division with temporal models is presented. To understand how bacteria migrate and proliferate to form complex multicellular behaviours such as biofilms, it is desirable to track individual bacteria and detect cell division events. Unlike eukaryotic cells, prokaryotic cells such as bacteria lack distinctive features, causing bacteria division difficult to detect in a single image frame. Furthermore, bacteria may detach, migrate close to other bacteria and may orientate themselves at an angle to the horizontal plane. Our system trains a hidden conditional random field (HCRF) model from tracked and aligned bacteria division sequences. The HCRF model classifies a set of image frames as division or otherwise. The performance of our HCRF model is compared with a Hidden Markov Model (HMM). The results show that a HCRF classifier outperforms a HMM classifier. From 2D bright field microscopy data, it is a challenge to separate individual bacteria and associate observations to tracks. Automatic detection of sequences with bacteria division will improve tracking accuracy.

An approach to jointly estimate 3D shapes and poses of stained nuclei from confocal microscopy images, using statistical prior information, is presented. Extracting nuclei boundaries from our experimental images of cell migration is challenging due to clustered nuclei and variations in their shapes. This issue is formulated as a maximum a posteriori estimation problem. By incorporating statistical prior models of 3D nuclei shapes into level set functions, the active contour evolutions applied on the images is constrained. A 3D alignment algorithm is developed to build the training databases and to match contours obtained from the images to them. To address the issue of aligning the model over multiple clustered nuclei, a watershed-like technique is used to detect and separate clustered regions prior to active contour evolution. Our method is tested on confocal images of endothelial cells in microfluidic devices, compared with existing approaches.

This paper presents an integrated stochastic approach to tracking multiple live cells that interact with the surrounding gel matrix. Cells migrate in a stochastic manner, forming a functional structure. Tracking the trajectory of each migrating cell and its interactions with the gel and other cells provide useful insights into how a vascular structure is formed as a collection of migratory cells. In micro-fluidic 3D angiogenic sprouting experiments, two types of images are obtained at discrete time steps using confocal microscopy: a) three-dimensional fluorescent images of stained cell nuclei and b) two dimensional visible light images of the gel matrix. These two sources of images provide supplementary information as the outline of the conduit or lumen formed in the matrix by the migrating cells can be seen in the images of the gel. A Bayesian filtering framework is developed which augments both the cell and conduit parameters to the same state vector, allowing mathematically consistent simultaneous observation updates from both channels. Issues encountered include the high dimensional state vector and non-Gaussian cell state updates. Results demonstrate that our method based on Rao-Blackwellized particle filtering treats these issues effectively.

This paper presents the development of an integrated mobile and static wireless sensor network (WSN) for the purpose of tracking a human in a home environment. Majority of tracking solutions rely on expensive sensors with high power requirements, which are not scalable nor suitable for systems requiring a long network lifetime. Our approach is to use inexpensive off-the-shelf stationary ranging sensors, placed at known locations in the environment. However, issues encountered when tracking a human with a limited number of sensors may include a poor tracking estimate and poor coverage regions where the static sensors are unable to detect the target of interest. An alternative to increasing the number of static sensors deployed, is to integrate single mobile sensor into the network. A collaborative sensor selection scheme is applied to static sensors which returns the measurement that will yield the highest information gain and maximise the network lifetime. Results show that with a robot following the human, the regions where the human is detected is increased and an improved tracking estimate is obtained from the fusion of static and mobile sensor readings.

Automation of biological cell tracking processes from time-lapse microscopy is increasingly being preferred due to time efficiency and avoidance of human inconsistencies. This paper presents development of algorithms to track the stochastic motion of the cells in micro-fluidic devices to aid the study of angiogenesis. Angiogenesis is the process of the growth of new blood vessels from a tissue monolayer. This process is important in several fields such as cancer treatment, tissue engineering and wound healing. Challenges with our data include low signal to noise ratio, complex cellular topologies such as close contact of cells and associating data over images acquired at different time. Our methodology of tracking of cells and the monolayer involves two steps; segmentation of the cells from images and tracking of these segmented cells over a span of time. The contribution of our work is the application of level set segmentation and the Bhattacharya distance measures for associating cells over sequential time frames. Results demonstrated on experimental data show that cells and monolayer are segmented and tracked effectively.

This paper describes a methodology for integrating confocal microscopy images acquired from two different illumination sources over time obtained from experiments of cell migration and vessel formation. The three-dimensional fluorescent images of stained cell nuclei and two dimensional bright field images of the extracellular matrix, supplement one another as the outline of the conduit formed by the migrating cells can be seen in the bright field images. A Bayesian filtering framework is developed to fuse the information in these two channels by creating and tracking a joint filter of nuclei and conduit parameters. This allows observation information from both channels to update both parameters simultaneously and the uncertainties of these parameters to be correlated providing a more accurate estimate.

This paper describes an automated visual tracking system combining time-lapse and end-point confocal microscopy to aid the interpretations of cell behaviors and interactions, with the focus on understanding the sprouting mechanism during angiogenesis. These multiple cells exhibit stochastic motion and are subjected to photo-bleaching and the images acquired are of low signal to noise ratio. Hence, following time-lapse imaging, high resolution end-point images are acquired. Our approach applies a probabilistic motion filter (a backward Kalman filtering followed by track smoothing) which incorporates end-point and all available time-lapse information in a mathematically consistent manner to obtain trajectory and phenotype information of multiple individual cells simultaneously. An extension of this algorithm, track smoothing with a Multiple Hypothesis Testing (MHT) data association, is proposed to improve association of multiple close contact and proliferating cells across images acquired from different time points to existing track trajectories. Our methodology was applied to tracking endothelial cell sprouting in three-dimensional micro-fluidic devices.

Applications for Wireless Sensor Networks may be decomposed into the deployment of tasks on different sensor nodes in the network. Task allocation algorithms assign these tasks to specific sensor nodes in the network for execution. Given the resource-constrained and distributed nature of Wireless Sensor Networks (WSNs), existing static (offline) task scheduling may not be practical. Therefore there is a need for an adaptive task allocation scheme that accounts for the characteristics of the WSN environment such as unexpected communication delay and node failure. In this paper, we focus on task allocation in WSNs which is performed with the aim of achieving a fair energy balance amongst the sensor nodes while minimizing delay using a market-based architecture. In this architecture, nodes are modeled as sellers communicating a deployment price for a task to the consumer. To address this task allocation problem, proposed price formulation is used as it continuously adapts to changes of the availabilities of resources. This scheme also accommodates for the node failure during task assignment. The Centralized and distributed message exchanged mechanisms between the nodes (sellers) and task allocator (consumer) are proposed to determine the winner among the sellers with the goal of reducing overhead and energy consumption. Simulation results show that, compared with a static scheduling scheme with an objective in energy balancing, the proposed scheme adapts to new environmental changes and uncertain network condition more dynamically and achieves a much better performance on energy balancing.

This paper introduces methods for allocating computational resources to track multiple targets in a sensor network of cameras, with the aim of maximizing tracking accuracy, an aspect of information quality (IQ). Particle filters are used where an independent filter is created for each target track. The particle filter's strength to track complex probability models comes at the cost of a higher demand for computational resources. As there is a direct correlation between the number of particles used and the computational costs, a means to meet the computational resource constraints is to limit the total number of particles. This motivates an online distribution of available particles to available tracks to maximise tracking accuracy. Restricting the number of particles may result in ineffective particle filters that diverge from the true target position. This issue has not been addressed in any known work of allocating particles. Our solution is a particle allocation scheme that allocates available particles to existing tracks based on both filter uncertainty and the effectiveness of the particle sample set. A method to allocate assigned particles in a track to overcome suspected divergence is also applied. Experimental results, from a multi-target tracking simulation, show that given the same number of particles to allocate, our scheme outperforms two other methods, which are (a) an equal (fixed) allocation of particles to each filter and (b) an allocation of particles solely based on filter uncertainty.

This paper presents the development and demonstration of non-Gaussian, decentralised state estimation using an outdoor sensor network consisting of an autonomous air vehicle, a manual ground vehicle, and two human operators. The location and appearance of landmarks were estimated using bearing only observations from monocular cameras. We show that inclusion of visual and identity information aids validation gating for data association when geometric information alone cannot discriminate individual landmarks. The combination of geometric, appearance, and identity information provided a common description (or map) of natural features for each of the nodes in the network. We also show the final map from the live demonstration which includes position estimates and classification labels of the observed features.

This paper presents a decentralised particle filtering algorithm that enables multiple vehicles to jointly track 3D features under limited communication bandwidth. This algorithm, applied within a decentralised data fusion (DDF) framework, deals with correlated estimation errors due to common past information when fusing two discrete particle sets. Our solution is to transform the particles into Gaussian mixture models (GMMs) for communication and fusion. Not only can decentralised fusion be approximated by GMMs, but this representation also provides summaries of the particle set. Less bandwidth per communication step is required to communicate a GMM than the particle set itself hence conversion to GMMs for communication is an advantage. Real airborne data is used to demonstrate the accuracy of our decentralised particle filtering algorithm for airborne tracking and mapping.

This paper presents two solutions for performing decentralised particle filtering in view of non-linear, non-Gaussian tracking in sensor networks. The issue is that no known methods exist to deal with correlated estimation errors due to common past information between two discrete particle sets. The first method transforms the particles to a Gaussian Mixture Model, the second approximates the set by a Parzen density estimate. Both of these representations accommodate consistent fusion and maintain accurate summaries of the particles. Requiring less bandwidth than particle representations, transformations to GMMs or Parzen representations for communication provide an added advantage. The accuracy in which the algorithms summarise the particle set, fusion methods and bandwidth requirements of each representation will be compared. Our results show that whilst less GMM components are required to summarise the sample statistics, the decentralised fusion solution using Parzen representations yields a more accurate result.

This paper presents a robust stochastic model for the incorporation of natural features within data fusion algorithms. The representation combines Isomap, a non-linear manifold learning algorithm, with Expectation Maximization, a statistical learning scheme. The representation is computed offline and results in a non-linear, non-Gaussian likelihood model relating visual observations such as color and texture to the underlying visual states. The likelihood model can be used online to instantiate likelihoods corresponding to observed visual features in real-time. The likelihoods are expressed as a Gaussian Mixture Model so as to permit convenient integration within existing nonlinear filtering algorithms. The resulting compactness of the representation is especially suitable to decentralized sensor networks. Real visual data consisting of natural imagery acquired from an Unmanned Aerial Vehicle is used to demonstrate the versatility of the feature representation.

Decentralised sensor networks typically consist of multiple processing nodes supporting one or more sensors. These nodes are interconnected via wireless communication. Practical applications of decentralised data fusion have generally been restricted to using Gaussian based approaches such as the Kalman or information filter. This paper proposes the use of Parzen window estimates as an alternate representation to perform decentralised data fusion. It is required that the common information between two nodes be removed from any received estimates before local data fusion may occur. Otherwise, estimates may become overconfident due to data incest. A closed form approximation to the division of two estimates is described to enable conservative assimilation of incoming information to a node in a decentralised data fusion network. A simple example of tracking a moving particle with Parzen density estimates is shown to demonstrate how this algorithm allows conservative assimilation of network information.