IRIS Research Staff

Yi Yao
Ph.D. Student

Office: 209 Ferris Hall
The University of Tennessee
Knoxville, TN 37996-2100
Telephone: (865) 974-9685
Fax: (865) 974-5459
E-mail: yyao1@utk.edu
Personal Web Page:  

 

Multiple camera surveillance- Multi-camera surveillance systems are introduced to achieve a continuous and robust object tracking by fulfilling the following three major functionalities: sensor planning, object tracking, and camera handoff.  For a continuous and automatic object tracking, a sufficient overlap of the camera’s field of view should be reserved so that camera handoff can be executed successfully before the object of interest becomes unidentifiable or untraceable.  We design a sensor planning algorithm that not only considers maximum coverage but also ensures sufficient overlapped camera’s field of view for an optimal handoff success rate.  This algorithm works in both static and dynamic environments using different types of cameras including static perspective, fisheye, and PTZ cameras.  Significantly improved handoff success rates are illustrated via experiments using both outdoor and indoor floor plans. 

 

Size preserving tracking- Conventional tracking concentrates on smooth and real-time pursuing of target motion.  Tracking with automatic zoom control abilities provides more flexibilities in target observing.  In addition to maintaining the target in the camera's field of view (FOV), the target image size is preserved so that target features are recognizable.  We look into video tracking algorithms which are capable of initializing proper zoom for desired target details and stabilizing the zoom for consistent target details.

 

Long range face recognition- Face image quality assessment and enhancement compensate for the performance degradations in face recognition rates caused by high system magnifications and long observation distances.  A class of adaptive sharpness measures is proposed to evaluate and predict this degradation.  A wavelet based enhancement algorithm with automated frame selection is developed and proves efficient via a considerably elevated face recognition rate for severely blurred long range face images.

 

 3D face reconstruction - In comparison with 2D face images, 3D face models have the advantage of being illumination and pose invariant, which provides improved capability of handling changing environments in practical surveillance.  Feature detection, as the initial process of reconstructing 3D face models from 2D uncalibrated image sequences, plays an important role and directly affects the accuracy and robustness of the resulting reconstruction.  In this paper, we propose an automated scene-specific selection algorithm that adaptively chooses an optimal feature detector according to the input image sequence for the purpose of 3D face reconstruction.  We compare the performance of various feature detectors in terms of accuracy and robustness of the sparse and dense reconstructions.  Our experimental results demonstrate the effectiveness of the proposed selection method from the observation that the chosen feature detector produces 3D reconstructed face models with superior accuracy and robustness to image noise.

 

High magnification imaging- Imaging systems with scopes are traditionally used in astronomy imaging and wide life observing.  More recently, the need for such capabilities has extended to wide area monitoring such as forest fires, harbors, and waterways.  In this project, we study a number of sensor arrangements for the acquisition of high magnification images.  The major sources of degradation resulted from high magnification are investigated and possible solutions are addressed.


Dual camera (PTZ & omnidirectional) video tracking systems- In multiple camera systems, PTZ camera is a popular choice.  The combination of omnidirectional and PTZ cameras explore their advantages and compensate for their weakness.  The dual camera system finds wide applications in indoor public area and hardware house monitoring.  Omnidirectional camera is responsible for latent motion detection while PTZ camera performs specific target pursing.