|
|
Research and Development
To achieve the objectives of this project, we shall first develop a robust face detector that can
work with most of the footage captured from typical surveillance cameras.
With a robust face detector, we can then research into 3D face reconstruction, and come up
with a method that enable us to map a 3D face model to each detected face. This 3D face
ITSP 6.0
model is important in the sense that it allows us to estimate the orientation of the face, which is
crucial for us to derive a transformation that reconstructs a frontal view of the detected face.
Super-resolution algorithm is then applied to the reconstructed face image to generate a better
and higher resolution image, on which the face recognition technology can work more reliably
and accurately.
Finally, we shall implement a people identification system prototype that takes typical video
surveillance cameras as inputs to illustrate the effectiveness of the technologies developed in
this project.
(I) Technologies to be developed
- Robust Face Detection Technology
This technology is related to the invention of detecting human faces from a flow of
image sequences, where the human face may be having different orientations.
Conventional face detection works well with the frontal face image only, the
technology we are going to developed intends to address this limitation.
- 3D Face Reconstruction
This technology is related to the invention of the 3D face model fitting to each
detected human face, so that the face/head orientations can be reliably estimated.
In addition, it is also a crucial technology for us to transform a non-frontal face
image back to a frontal one so that the reconstructed face image can work better
with COTS face recognition packages.
- Face Image SuperResolution
This technology is related to the invention of the method that performs
supersampling on multiple face images to generate better resolution image for face
recognition purposes.
(II) Innovation use of existing technologies
As the objective of this project is to explore more application scenario for COTS
face recognition package, one of the technologies we are going to use is the face
recognition technology.
Apart from face recognition technology, we shall also innovatively employ
state-of-the-art image processing and computer vision technologies for the
development of the new technologies stated in (I). That includes haar-feature
detectors for face detection, shape-based model fitting for face orientation
estimation, and 3D reconstruction techniques for face reconstruction.
|
|
|
|
|
