In this study, we suggest and validate an all-numerical implementation of a VanderLugt correlator which is optimized for
face recognition applications. The main goal of this implementation is to take advantage of the benefits (detection,
localization, and identification of a target object within a scene) of correlation methods and exploit the reconfigurability
of numerical approaches. This technique requires a numerical implementation of the optical Fourier transform. We pay
special attention to adapt the correlation filter to this numerical implementation. One main goal of this work is to reduce
the size of the filter in order to decrease the memory space required for real time applications. To fulfil this requirement,
we code the reference images with 8 bits and study the effect of this coding on the performances of several composite
filters (phase-only filter, binary phase-only filter). The saturation effect has for effect to decrease the performances of the
correlator for making a decision when filters contain up to nine references. Further, an optimization is proposed based for
an optimized segmented composite filter. Based on this approach, we present tests with different faces demonstrating
that the above mentioned saturation effect is significantly reduced while minimizing the size of the learning data base.
An optimized technique, based on the fringe-adjusted joint transform correlator architecture, is proposed and validated for rotation invariant recognition and tracking of a target in an unknown input scene. To enhance the robustness of the proposed technique, we used a three-step optimization. First, we utilized the fringe-adjusted filter (H FAF ) in the Fourier plane, then we added nonlinear processing in the Fourier plane, and, finally, we used a new decision criterion in the correlation plane by considering the correlation peak energy and the highest peaks outside the desired correlation peak. Several tests were conducted to reduce the number of reference images needed for fast tracking, while ensuring robust discrimination and efficient tracking of the desired target. Test results, obtained using the pointing head pose image database, confirm robust performance of the proposed method for face recognition and tracking applications. Thereafter, we also tested the proposed technique for a challenging application such as underwater mine detection and excellent results were obtained.
Interestingly, the past 20 years have provided us many examples of optical correlation methods for pattern recognition,
e.g. VanderLugt correlator (VLC). In recent years, hybrid techniques, i.e. numerical implementation of correlation, have
been also considered an alternative to all-optical methods because they show a good compromise between performance
and simplicity. Moreover, these correlation methods can be implemented using an all-numerical and reprogrammable
target such as the graphics processor unit (GPU), or the field-programmable gate array (FPGA). However, this numerical
procedure requires realizing two Fourier Transforms (FT), a spectral multiplication, and a correlation plane analysis. The
purpose of this study is to compare the performances of a numerical correlator based on the fast Fourier transform (FFT)
with that relying on a simulation of the Fraunhofer diffraction. Different tests using the Pointing Head Pose Image
Database (PHPID) and considering faces with vertical and horizontal rotations were performed with the code MATLAB.
Tests were conducted with a five reference optimized composite filter. The receiving operating characteristics (ROC)
curves show that the optical FT simulating the Fraunhofer diffraction leads to better performances than the FFT. The
implications of our results for correlation are discussed.
In this paper, we propose a new technique for rotation invariant recognition and tracking of the face of a target person in
a given scene. We propose an optimized method for face tracking based on the Fringe-adjusted JTC architecture. To
validate our approach, we used the PHPID data base containing1 faces with various in-plane rotations. To enhance the
robustness of the proposed method, we used a three-step optimization technique by: (1) utilizing the fringe-adjusted filter
(HFAF) in the Fourier plane, (2) adding nonlinearity in the Fourier plane after applying the HFAF filter, and (3) using a new
decision criterion in the correlation plane by considering the correlation peak energy and five largest peaks outside the
highest correlation peak. Several tests were made to reduce the number of reference images needed for fast tracking
while ensuring robust discrimination and efficient of the desired target.
In this paper, we explore the use of optical correlation-based recognition to identify and position underwater
man-made objects (e.g. mines). Correlation techniques can be defined as a simple comparison between an
observed image (image to recognize) and a reference image; they can be achieved extremely fast. The result
of this comparison is a more or less intense correlation peak, depending on the resemblance degree between
the observed image and a reference image coming from a database. However, to perform a good correlation
decision, we should compare our observed image with a huge database of references, covering all the appearances
of objects we search. Introducing all the appearances of objects can influence speed and/or recognition quality.
To overcome this limitation, we propose to use composite filter techniques, which allow the fusion of several
references and drastically reduce the number of needed comparisons to identify observed images. These recent
techniques have not yet been exploited in the underwater context. In addition, they allow for integrating some
preprocessing directly in the correlation filter manufacturing step to enhance the visibility of objects. Applying
all the preprocessing in one step reduces the processing by avoiding unnecessary Fourier transforms and their
inverse operation. We want to obtain filters that are independent from all noises and contrast problems found
in underwater videos. To achieve this and to create a database containing all scales and viewpoints, we use as
references 3D computer-generated images.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.