The use of computational mechanisms, more precisely those that approach the concepts
of computer vision, are of great relevance being widely applied in areas such as biometric
identification, detection and recognition of objects. Some works show these applications
as in the identification of faces (CHE; CHANG, 2010) and eyes (MANNAY; ABID, 2015).
For the processing of the images, devices that have high computational power are required,
thus the development of embedded systems using the reconfigurable computing paradigm
(RC), such as the FPGA (Field Programmable Gate Array), is promising in this context.
Unlike other approaches such as Application Specific Integrated Circuits (ASICs) where
hardware is developed that can not subsequently be reprogrammed, FPGAs provide flexibility
in the development of specific functionalities. Therefore, when future design updates
and upgrades to new technologies are required, it offers the developer the option of synthesizing
a new hardware description (HDL). In the face of the advantages offered by FPGAs,
we adopted the same with the objective of performing the facial detection process, and
opted for the use of a customizable Intel/Altera Nios II processor, SDRAM (Synchronous
Dynamic Random Access Memory) and other components necessary for the operation of
the system, which will be responsible for storing and executing the algorithm responsible
for said detection. We use the algorithm proposed by (VIOLA; JONES, 2001), developed
in C/C++, to perform the image processing and indicate the presence or absence of one
or more faces. After the system development and integration of the algorithm, executions
were performed with the reading of images previously stored in the computer and transmitted
to the FPGA and also the capture from the Terasic TRDB-D5M camera. It was
observed the behavior of face detections in several images with characteristics such as face
in normal position, inclined, with and without the use of glasses, different formats of face
and hair. We succeeded in executing the scenarios presented, as well as the detection and
demarcation of the faces locations.
|
