Category: Tracking

ACVT conducts research in the fields of camera calibration, visual tracking, novel-view synthesis, automatic visual surveillance, and simultaneous localisation and mapping. This area is led by Laureate Professor Ian Reid. Visual SLAM. By observing a static scene with a camera, it is possible to estimate the trajectory of the camera and simultaneously build a map of the environment. When done in real time this has the potential to turn a simple webcam into a flexible and powerful geometric sensing device. The area has had a major impact in the robotics community because for the first time a visual sensor could deliver timely geometric information about an unstructured environment. Visual Geometry. Advances in the field of visual geometry based upon an understanding of projective geometry as applied to making physical geometric measurements from uncalibrated images. Camera Self-Calibration. Calibration of cameras using invariant eigenvectors from projective transformations, mapping scene structure from one location to another. Also, work has been done in methods for self-calibration of rotating and zooming camera. Stereo Reconstruction. Use of novel graph construction methods that permit second-order interactions to build piecewise planar reconstructions. Human motion capture. Development of the first markerless motion capture system. Through the deployment of annealed particle filtering and extensions to incorporate a genetic “crossover” operator that enabled efficient exploration of the huge multi-modal configuration space. Visual tracking and active vision. The ability to track an object of interest is a fundamental enabling technology for many video analysis tasks. A distinguishing feature of the area has been an emphasis on real-time methods, and incorporating these into robotic pan–tilt–zoom platforms for closed-loop visual tracking. The significant processing power of a modern desktop computer has permitted greater sophistication in visual tracking, and recent work has shown how a target can be segmented from its background while being tracked in real time. A further development in computing hardware has led to work in tracking human heads in surveillance video. Visual surveillance and activity recognition. Visual tracking combined with our work on SLAM in radar yielded a system for marine situation awareness that tracks targets in radar, self-localises without the need for GPS, and drives a custom high-performance pan–tilt–zoom platform to gather visual data about any designated target.

Visual tracking of multiple objects: A stochastic geometrical approach (DP0880553)

Prof. David Suter (Then of Monash University) along with Associate Prof. B Vo has been awarded a 3 year ARC Discovery Grant valued at $235,000. Reliable real-time visual multiple-object tracking techniques will open up new applications that enhance the quality of life such as driving safety, traffic monitoring, home security, security and surveillance of public […]

Posted in Parameter Estimation, Projects, Research, Surveillance, Tracking | Tagged , , , |

Comments Off on Visual tracking of multiple objects: A stochastic geometrical approach (DP0880553)