I Face Detection And Recognition System author: Mukund Agarwal supervisor: Professor Nishan Canagarajah Project Thesis submitted in support of the Degree of Bachelor of Engineering in Electronic and Communications Engineering

Figure 10 Cascaded classifier 
Figure 8 Basic Haar wavelets 

---

Mukund Agarwal 
Face Detection & Recognition System 
12 
4 . 2 I m p l e m e n t a t i o n
After researching, a MATLAB® executable (MEX) 
implementation of ‘Voila and Jones’ method was 
found on MATLAB® Central website. Developed by 
Sreekar Krishna
9
, the code uses Haar classifier 
provided by Intel for Open Source Computer Vision 
Library (Open CV). 
We only run the face detection and the other module 
once motion is detected. If there is no motion then 
there is no point in running them as it will just cause 
unnecessary processing load. The argument here is 
that if someone came in the scene then there was an 
initial motion and those modules have already 
processed that image once. The gray scale image used 
in motion detection is passed to the face detection 
function which is written in C++. The function then 
returns the number of faces detected and their 
respective coordinates. 
A bug was found in the downloaded code. When 
multiple faces were detected the passed array from 
the C++ code to MATLAB® environment wasn’t in the 
right format as expected. This was corrected by 
changing the format of the array. 
Once detection coordinates were acquired, the 
original coloured image is displayed and the locations 
of the faces are marked on the picture.
If there are any faces detected then those coordinates 
and the image are passed to the face recognition 
module. Else we loop back to the motion detection. 
Figure 11 shows the flowchart which is implemented 
in the code.
 
Figure 11 Face Detection Algorithm
 

---

Mukund Agarwal 
Face Detection & Recognition System