Xunyi Yu

DIORAMA: Dynamic information collection and resource tracking architecture

DIORAMA is a real-time scalable decision support framework built on rapid information collection and accurate resource tracking functionalities. Using RFID technology the proposed system tracks emergency responders and victims at the disaster scene. DIORAMA improves the accuracy and decreases the time it takes rescuers to triage, treat and evacuate victims from a disaster scene, as compared to the traditional methods and process that involves using paper triage tags. The information can then be viewed from a website that shows a satellite image of the disaster area with icons representing the paramedics and victims.

MiRTE: Mixed Reality Triage and Evacuation game for Mass Casualty information systems design, testing and training

In this paper we introduce a Mixed Reality Triage and Evacuation game, MiRTE, that is used in the development, testing and training of Mass Casualty Incident (MCI) information systems for first responders. Using the Source game engine from Valve software, MiRTE creates immersive virtual environments to simulate various incident scenarios, and enables interactions between multiple players/first responders. What distinguishes it from a pure computer simulation game is that it can interface with external mass casualty incident management systems, such as DIORAMA. The game will enable system developers to specify technical requirements of underlying technology, and test different alternatives of design. After the information system hardware and software are completed, the game can simulate various algorithms such as localization technologies, and interface with an actual user interface on PCs and Smartphones. We implemented and tested the game with the DIORAMA system.

Global Identification of Tracklets in Video Using Long Range Identity Sensors

Reliable tracking of people in video and recovering theiridentities are of great importance to video analytics applications.For outdoor applications, long range identity sensorssuch as active RFID can provide good coverage in alarge open space, though they only provide coarse locationinformation. We propose a probabilistic approach usingnoisy inputs from multiple long range identity sensorsto globally associate and identify fragmented tracklets generatedby video tracking algorithms. We extend a networkflow based data association model to recover tracklet identityefficiently. Our approach is evaluated using five minutesof video and active RFID measurements capturing four peoplewearing RFID tags and a couple of passersby. Simulationis then used to evaluate performance for larger numberof targets under different scenarios.identities are of great importance to video analytics applications.For outdoor applications, long range identity sensorssuch as active RFID can provide good coverage in alarge open space, though they only provide coarse locationinformation. We propose a probabilistic approach usingnoisy inputs from multiple long range identity sensorsto globally associate and identify fragmented tracklets generatedby video tracking algorithms. We extend a networkflow based data association model to recover tracklet identityefficiently. Our approach is evaluated using five minutesof video and active RFID measurements capturing four peoplewearing RFID tags and a couple of passersby. Simulationis then used to evaluate performance for larger numberof targets under different scenarios.

Scalable patients tracking framework for mass casualty incidents

We introduce a system that tracks patients in a Mass Casualty Incident (MCI) using active RFID triage tags and mobile anchor points (DM-tracks) carried by the paramedics. The system does not involve any fixed deployment of the localization devices while maintaining a low cost triage tag. The localization accuracy is comparable to GPS systems without incurring the cost of providing a GPS based device to every patient in the disaster scene.

Audible vision for the blind and visually impaired in indoor open spaces

In this paper we introduce Audible Vision, a system that can help blind and visually impaired users navigate in large indoor open spaces. The system uses computer vision to estimate the location and orientation of the user, and enables the user to perceive his/her relative position to a landmark through 3D audio. Testing shows that Audible Vision can work reliably in real-life ever-changing environment crowded with people.

Mass casualty incident surveillance and monitoring using identity aware video analytics

In this paper, we propose an identity aware video analytic system that can assist securing the perimeter of a mass casualty incident scene and generate identity annotated video records for forensics and training purposes. Establishing a secure incident scene perimeter and enforcing access control to different zones is a demanding task for current video surveillance systems which lack the ability to provide the identity of the target and its security clearance. Our system which combines active RFID sensors with video analytic tools recovers the identity of the target enabling the activation of suitable alert policies. The system also enables annotation of incident scene video with identity metadata, facilitating the incident response process reconstruction for forensics analysis and emergency response training.

RiBOMS: RFID-based object management system for home environments

In this paper we introduce a RFID-based object management system, RiBOMS, for home environments. The system has an easy to use pictorial user interface aimed at older adults with associative memory impairments. The system technical correctness was successfully tested in a lab environment.

A calibration free hybrid RF and video surveillance system for reliable tracking and identification

We introduce a calibration free hybrid RF and video surveillance system for reliable tracking, identification in Mass Casualty Incidents (MCI). The system fuses RF measurements along with information harvested from the video such as target temporal, kinematic, and appearance information. Our hybrid RF and video surveillance system can be deployed rapidly without manual calibration of either the camera or the RF devices, and can work in complex RF environments. This is made possible by the simultaneous identification and mapping algorithm that iteratively refines the mapping of the radio beacon signal strength signature to target positions in video frames. The system is the basis of identity aware video surveillance, i.e. enabling unique identification of targets in the video and applying identity specific analytics roles to individual targets.

Detecting and identifying people in mobile videos

In this paper, we propose a system capable of detecting and identifying people in videos captured by smart phones. We discuss the challenges to extend existing location aware multimedia applications from annotating static landmarks in distance to annotating dynamic people in close range with significant pose variations. We propose to use a hybrid video and RF tracking system to enable accurate observer and target localization, and extract part models comprised of Maximally Stable Color Regions for each target. The model can efficiently detect possible positions of targets in the video, which are then used as dynamic landmarks to calibrate the camera orientation. Finally, positions of all targets in the video are jointly estimated using both visual features and spatial constraints. Experiments show that our approach can locate identified targets in video with significantly higher accuracy than back-projection using camera orientation estimations from accelerometers and magnetometers.