Advisor: Dr. Chris Pollett

Description:

In this CS297 project, we intend to describe a potential technique of helping to identify missing children. This technique of using a pattern classifier to identify images of missing children is based on principal components analysis (Eigenface approach). Given that digital images are inherently high in dimensionality, Principal Components Analysis is useful in that it attempts to project high dimensional data onto a low dimensional space in a least squares fashion. This has the advantage of relatively low requirements on memory and computation. This low dimensional space may then be used for the purposes of image classification or reconstruction. Input data to the classifier will be in the form of digital images from a website for missing children. The images of missing children will come in two sets, where one set is of the original images, and the other set that is age-progressed (artificially aged pictures by digital enhancements) as retrieved. The low dimensional space of both sets of images form the basis vectors for image recognition as well as age-progressed prediction of each child. In other words, given an image of a missing child, the classifier not only attempts to identify the child, but produces a predicted image of an older child.

Schedule:

Deliverables:

The full project will be done when CS298 is completed. The following will be done by the end of CS297:

1. A program that implements the Eigenface approach, as a proof of concept, to classify gray-scale images of human faces using the following two factors: image is or is not a face, and image is or is not recognized. Therefore, there are a total of four groupings, e.g. image is a face and is recognized. It is assumed that the images will be cropped to the right size and scale, and normalized in terms of orientation and lighting.

2. Python scripts that enters a website for missing children, downloads all relevant images by using the web-page metadata, preprocesses the images in terms of size, scale, and orientation, and sets up an appropriate directory hierarchy to organize and store the images.

3. Revised program of deliverable 1 that attempts to use different statistical models to improve on image recognition performance.

4. Revised program of deliverable 3 that attempts to perform age progression given a new image of a missing child.

5. The final CS 297 report will be completed.

References:

[Bartlett02] Face Recognition by Independent Component Analysis. Marian Stewart Bartlett, Javier R. Movellan, and Terrence J. Sejnowski. IEEE Transactions on Neural Networks, Vol. 13, No. 6, November 2002.

[Belhumeur96] Eigenfaces Vs. Fisherfaces: Recognition Using Class Specific Linear Projection. Peter N. Belhumeur, Joao P. Hespanha, and David J. Kriegman. European Conference on Computer Vision, 1996.

[Callender01] Perl for Web Site Management. John Callender. O’Reilly, 2001.

[Duda01] Pattern Classification. Richard O. Duda, Peter E. Hart, and David G. Stork. John Wiley & Sons, 2001.

[Gonzalez02] Digital Image Processing. Rafael C. Gonzalez and Richard E. Woods. Prentice-Hall, Inc., 2002.

[Mitchell97] Machine Learning. Tom M. Mitchell. McGraw Hill, 1997.

[Moghaddam98] Beyond Eigenfaces: Probabilistic Matching for Face Recognition. Baback Moghaddam, Wasiuddin Wahid, and Alex Pentland. 3rd IEEE International Conference on Automatic Face & Gesture Recognition, April 1998.

[Pentland94] View-Based and Modular Eigenspaces for Face Recognition. Alex Pentland, Baback Moghaddam, and Thad Starner. IEEE Conference on Computer Vision & Pattern Recognition, 1994.

[Smith02] A Tutorial on Principal Components Analysis. Lindsay I. Smith. http://www.cs.otago.ac.nz/cosc453/student_tutorials/principal_components.pdf, 2002.

[Turk91] Eigenfaces for Recognition. Matthew Turk and Alex Pentland. Journal of Cognitive Neuroscience, Vol. 3, No. 1, 1991.