A5064997605- Target Tracking and Data Fusion in Sensor Networks
- Distributed Sensor Networks and Detection Algorithms
- Infrared Target Detection Methodologies
- Fault Detection and Control Systems
- Video Surveillance and Tracking Methods
- Anomaly Detection Techniques and Applications
- Gaussian Processes and Bayesian Inference
- Robotics and Sensor-Based Localization
- Guidance and Control Systems
- Control Systems and Identification
- Advanced Vision and Imaging
- 3D Surveying and Cultural Heritage
- Retinal Imaging and Analysis
- Water Systems and Optimization
- Autonomous Vehicle Technology and Safety
- Industrial Vision Systems and Defect Detection
- Network Security and Intrusion Detection
- Image and Object Detection Techniques
- Neural Networks and Applications
- Adversarial Robustness in Machine Learning
- Music and Audio Processing
- Energy Efficient Wireless Sensor Networks
- Distributed Control Multi-Agent Systems
- Retinal Diseases and Treatments
- AI in cancer detection
RMIT University
2015-2024
MIT University
2021-2024
The Royal Melbourne Hospital
2022
Institute of Engineering
2017
Retinal swelling due to the accumulation of fluid is associated with most vision-threatening retinal diseases. Optical coherence tomography (OCT) current standard care in assessing presence and quantity image-guided treatment management. Deep learning methods have made their impact across medical imaging, many OCT analysis been proposed. However, it currently not clear how successful they are interpreting on OCT, which lack standardized benchmarks. To address this, we organized a challenge...
Surface crack detection is an integral part of infrastructure health surveys. This work presents a transformative shift towards rapid and reliable data collection capabilities, dramatically reducing the time spent on inspecting infrastructures. Two unmanned aerial vehicles (UAVs) were deployed, enabling capturing images simultaneously for efficient coverage structure. The suggested drone hardware especially suitable inspection with confined spaces that UAVs broader footprint are incapable...
This letter addresses the sensor selection problem for tracking multiple dynamic targets within a network. Since bandwidth and energy of network are constrained, it would not be feasible to directly use entire information nodes detection hence need selection. Our solution is formulated using multi-Bernoulli random finite set framework. The proposed method selects minimum subset sensors which most likely provide reliable measurements. overall scheme robust that works in challenging scenarios...
This paper presents a sensor-control method for choosing the best next state of sensors that provide accurate estimation results in multitarget tracking application. The proposed solution is formulated multi-Bernoulli filter and works via minimization new estimation-error-based cost function. Simulation demonstrate can outperform state-of-the-art methods terms computation time robustness to clutter while delivering similar accuracy.
This paper proposes a novel method in order to obtain voxel-level segmentation for three fluid lesion types (IR-F/SRF/PED) OCT images provided by the ReTOUCH challenge [1]. The is based on deep neural network consisting of encoding and de-coding blocks connected with skip-connections which was trained using combined cost function comprising cross-entropy, dice adversarial loss terms. results held-out validation set shows that architecture functions used has resulted improved retinal...
In this paper, we propose an efficient approach for industrial defect detection that is modeled based on anomaly using point pattern data. Most recent works use \textit{global features} feature extraction to summarize image content. However, global features are not robust against lighting and viewpoint changes do describe the image's geometrical information be fully utilized in manufacturing industry. To best of our knowledge, first transfer learning local/point overcome these limitations...
A constrained sensor control method is presented for multiobject tracking using labeled multi-Bernoulli filters. The proposed framework based on a novel approximation of the Cauchy-Schwarz divergence between prior and posterior densities, which does not need Monte Carlo sampling random sets in space. void probability functional also formulated distributions used within our to form solution. Numerical studies demonstrate that reasonably acceptable movements are decided controlled by method,...
This paper presents a new sensor management method for multitarget filtering, that is designed based on maximizing measure of confidence in accuracy the state estimate. Confidence estimation quantified by optimal subpattern assignment-based dispersion posterior about its statistical mean. Implementation algorithm generic filters presented. Simulation studies with labeled multi-Bernoulli filter demonstrate excellent performance challenging control scenarios.
This paper presents a new approach towards statistical fusion of multi-source information. Our solution is formulated in the context fusing Poisson finite random set posteriors returned by multiple local PHD filters at sensor nodes distributed multi-sensor multi-object estimation system. The most common measure used for information gain stochastic Kullback-Leibler divergence (KLD) which leads to well-known Generalised Covariance Intersection (GCI) rule fusion. We present idea using...
This paper presents a new solution for multi-target tracking over network of sensors with limited spatial coverage. The proposed is based on the centralized data fusion architecture. main contribution introduction track-to-track approach in which posterior distributions states, reported by various sensor nodes, are fused way that redundant information combined and rest complement each other. formulated within labeled random finite set framework incorporates all state label provided multiple...
A new approach to solve the sensor control problem is proposed, formulated based on multi-object Bayes filtering in partially observable Markov decision process (POMDP) context, where states are assumed be random finite sets with multi-Bernoulli distributions. We introduce a novel cost function that reliable real-time environment. In each iteration, after predicting parameters, estimates for number and of targets extracted. For admissible command, Monte-Carlo samples measurements...
With the recent advent of labeled random finite set filters, it is now possible to not only estimate number objects and their states, but also track trajectories all within stochastic filtering scheme. This paper investigates how label information returned by a multi-Bernoulli filter can be effectively used for sensor control purposes. The main focus on selective multi-object tracking applications where with particular labels are high priority, needs controlled achieve maximum confidence in...
Tracking multiple objects through time is an important part of intelligent transportation system. Random finite set (RFS)-based filters are one the emerging techniques for tracking objects. In multi-object (MOT), a common assumption that each object moving independent its surroundings. However, in many real-world applications, interact with another and environment. Such interactions rarely considered within process. this paper, we present novel approach to incorporate target prediction step...
The recently developed labeled multi-Bernoulli (LMB) filter uses better approximations in its update step, compared to the unlabeled filters, and more importantly, it provides us with not only estimates for number of targets their states, but also labels existing tracks. This paper presents a novel sensor-control method be used optimal multi-target tracking within LMB filter. proposed task-driven cost function which both state estimation errors cardinality are taken into consideration....
One of the core challenges in visual multi-target tracking is occlusion. This especially important applications such as video surveillance and sports analytics. While offline batch processing algorithms can utilise future measurements to handle occlusion effectively, online have rely on current past only. As such, it markedly more challenging applications. To address this problem, we propagate information over time a way that generates sense déjà vu when similar motion features are observed....