Thesis. Facial recognition security system

Facial recognition security system

This report is being presented as a partial fulfillment in the requirement of the
degree as a Bachelor of Computer Engineering
At

COMSATS Institute of Information Technology Abbottabad, Pakistan
(Session:2009 - 2013)
Submitted by:
Tallah jamshaid

Fa09-bce-051

Hashim khan

sp08-bee-153

Israr-ul-haq

sp08-bce-020

Supervised by: Eng. Nauman Tareen
Approved by :

Department of Electrical Engineering
COMSATS Institute of Information Technology Abbottabad

1

In the Name of ALLAH
The Most Gracious
The Most Merciful

Yesterday ALLAH was with us,
Today we are under HIS persone
&
We should not worried about tomorrow,
ALLAH is already there
We just need to ask for HIS Mercy to all.

2

Acknowledgement

Almighty ALLAH is very kind, merciful and compassionate .His benevolence and
blessings enabled us to accomplish this task. I thank Almighty ALLAH the most
beneficent, the most merciful. I offer my humblest gratitude from deepest core of heart to
the Holy Prophet “Hazarat Muhammad” (Peace Be Upon Him), who is forever will be a
model of guidance and knowledge for humanity as a whole.

We had dedicated this project to our Parents and Family, whose love and affection had
been inspirational throughout our lives,
To our Teachers to whom we owe a lot for success in our personeer
AND
To all those who in any capacity had been help full in the project.

We are thankful to our project supervisor Engr. Nauman Tareen, lecturer CIIT/EE
COMSATS University, Abbottabad. It is hard to find words of appropriate dimensions to
express our gratitude to our worthy supervisor for his keen potential interest, suggestions,
consistent encouragement and support throughout the course of this project.

We are highly grateful to Engr. Asmat ali shah without whose help and guidance, this
project could not have been a success.

3

Abstract
A comprehensive research has been undertaken to design and develop a system that will
count ,detect and measure speed of Faces for highways .This project describes a visual
object detection framework that is capable of processing images extremely rapidly while
achieving high detection rates ,will count detected object and will measure speed of
detected object. There are five key contributions.
The first is the introduction of a new image representation called the “Integral Image”
which allows the features used by our detector to be computed very quickly. The second
is a learning algorithm, based on AdaBoost, which selects a small number of critical
visual features and yields extremely efficient classifiers. The third contribution is a
method for combining classifiers in a “cascade” which allows background regions of the
image to be quickly dispersonded while spending more computation on promising objectlike regions. The fourth and fifth contribution is to count Faces and measure speed of
Faces by processing on pixels of an image.
Facial recognition security system has a wide range of applications, such as it is useful
for security system in many sensitive areas and helpful to reduces security risks in many
departments and in institutes.
A set of experiments in the domain of facial recognition security system presented. The
system yields Facial recognition security system performance comparable to the best
previous systems.

Contents
Table of Content:
4

Chapter 1

Introduction ......................................................................................... 10

1.1 facial Detection: ................................................................................................................................ 10
1.2 data base: ............................................................................................. Error! Bookmark not defined.
1.3 microcontroller: ................................................................................... Error! Bookmark not defined.
1.4Objective: .............................................................................................. Error! Bookmark not defined.
Chapter 2

Image Processing ............................................................................................................ 13

2.1 Background: ...................................................................................................................................... 13
2.2 Types of Processing: .......................................................................................................................... 13
2.2.1 Low-level processing: ................................................................................................................. 14
2.2.2 Mid-level processing: ................................................................................................................. 14
2.2.3 Higher-level processing: ............................................................................................................. 14
2.3 Components of an Image Processing System: .................................................................................. 14
2.4 Applications: ..................................................................................................................................... 16
CHAPTER 3

Face Detection ............................................................................................................ 18

3.1 Introduction ....................................................................................................................................... 18
3.2 Rectangular features (Haar features): .............................................................................................. 18
3.3 An Integral Image for rapid feature computation: ......................................................................... 20
3.4 The AdaBoost Machine –learning method: ...................................................................................... 22
3.4.1 Working of Adaboost Algorithm: ............................................................................................... 23
3.5 The Cascaded Classifier: .................................................................................................................... 24
3.5.1 Training a cascade: ..................................................................................................................... 25
3.5.2 Optimum Combination: ............................................................................................................. 26
CHAPTER 4

Database ......................................................... Error! Bookmark not defined.

4.1 Types of database: ............................................................................................................................ 19
4.2 :Extensible Markup Language Database .............................................. Error! Bookmark not defined.
4.3 Storing Image in XML database ........................................................... Error! Bookmark not defined.
4.3.1 : ..................................................................................................... Error! Bookmark not defined.
4.3.2 Feature count technologies: ......................................................... Error! Bookmark not defined.
4.4 Applications to collect Feature data .................................................... Error! Bookmark not defined.
5.3 Image Processing: ................................................................................ Error! Bookmark not defined.
Chapter 6

Conclusion.................................................................................................. 29
5

6.1 DETECTION: ....................................................................................................................................... 29
Appendix A .................................................................................................................................................. 30
Appendix B .................................................................................................................................................. 38
APPENDIX C ................................................................................................................................................. 41

Table of Figures:

Figure 2.1 Components of general purpose image processing system. ................................................. 15
Figure.3-1. Example rectangle features shown relative to the enclosing detection window. The sum of
the pixels which lie within the white rectangles are subtracted from the sum of pixels in the grey
rectangles. Two-rectangle features are shown in (A) and (B). Figure (C) shows a three-rectangle
feature, and (D) a four-rectangle feature. ................................................................................................ 19
Figure 3-2. shows how an integral image formed from the input image……………………………………18
Figure 3-3. The value of the integral image at point (x, y) is the sum of all the pixels above and to the
left………………………………………………………………………………….……………………………….19
Figure.3-4.The sum of the pixels within rectangle D can be computed with four array references. The value of
the integral image at location 1 is the sum of the pixels in rectangle A. The value at location 2 is A + B, at
location 3 is A + C, and at location 4 is A + B + C + D. The sum within D can be computed as 4 + 1 − (2 +
3)…………………………………………………………………………………………………………………..19

6

Figure. 3.5 This feature is selected by AdaBoost. The feature measures the difference in intensity
between the region below headfeatures and a region of shadow between the tyres. The feature
capitalizes on the observation that the headfeature region is often featureer than the shadow region. 25
Figure. 3.6. Schematic depiction of a detection cascade. A series of classifiers are applied to every subwindow. The initial classifier eliminates a large number of negative examples with very little processing.
Subsequent layers eliminate additional negatives but require additional computation. After several stages of
processing the numbers of sub-windows have been reduced radically. Further processing can take any form
such as additional stages of the cascade (as in our detection system) or an alternative detection system……25
Figure. 3.7. Positive Samples used in training process…………………………………………………………..26
Figure. 3.7. Positive Samples used in training process………………………………………………………26
Figure. 3.8. Negative Samples used in training process…………………………………………………………27
Figure 4.1 shows the counting of Faces on the Detection
Window…………………………………………………………29
Figure 4.2 shows different technologies of Face count…………………………………………………………………33
Figure 6.1 shows Face detection…………………………………………………………………………………41
Figure 6.2 shows the Face count………………………………………………………………………………42
Figure 6.3 shows the Face speed……………………………………………………………………………..43

7

8

9

Chapter 1

Introduction

1.1 Biometric Security Systems
Biometrics is used in the process of authentication of a person by verifying or Identifying that a user
requesting a network resource is who he, she, or it claims to be, and vice versa. It uses the property
that a human trait associated with a person itself like structure of finger, face details etc. By
comparing the existing data with the incoming data we can verify the identity of a particular person.
There are many types of biometric system like fingerprint recognition, face detection and
recognition, iris recognition etc., these traits are used for human identification in surveillance system,
criminal identification. Advantages of using these traits for identification are that they cannot be
forgotten or lost. These are unique features of a human being which is being used widely.

1.2Goal
The goals of this project are to enhance public safety, reduce congestion, improved travel and transit
information, generate cost savings to person personriers and emergencies operators, reduce
detrimental environmental impacts, etc. This technology assists states, cities, and towns nationwide
to meet fully the increasing demands on Facial recognition Security system. The efficiency of the
system is mainly there is based on the performance and comprehensiveness of the Face detection
technology. Face on detection and tracking are an integral part of any Face detection technology,
since it gathers all or part of the information that are used in an efficient way.

1.3 Face Detection:
In recognition, Face detection system may be defined as a system which is capable of detecting faces
and measure Feature parameters such as count, speed, incidents, etc. Face detection by video cameras
is one of the most promising non-intrusive technologies for large-scale data collection and
implementation of advanced Feature control and management schemes. Face detection is also the
basis for Face tracking. The correct Face detection results in better tracking. Modern computer
controlled Feature systems have more complex Face detection requirements than those adopted for
normal Feature-actuated controllers for Feature signals, for which many off-the-shelf Face detectors
were designed. Many useful and comprehensive parameters like count, speed, Face classification,
queue lengths, volume/lane, lane changes, microscopic and macroscopic behaviors can be evaluated
through video based Face detection and tracking.

1.4 Face Recognition
Face is a complex multidimensional structure and needs good computing techniques for recognition.
The face is our primary and first focus of attention in social life playing an important role in identity of
individual. We can recognize a number of faces learned throughout our lifespan and identify that faces at
a glance even after years. There may be variations in faces due to aging and distractions like beard, glasses
or change of hairstyles. Face recognition is an integral part of biometrics. In biometrics basic traits of
human is matched to the existing data and depending on result of matching identification of a human
being is traced. Facial features are extracted and implemented through algorithms which are efficient
and some modifications are done to improve the existing algorithm models. Computers that detect and
recognize.

10

Faces Could Be Applied to A Wide variety of practical applications including criminal identification,
security systems, identity verification etc. Face detection and recognition is used in many places now a
days , in websites hosting images and social networking sites. Face recognition and detection can be
achieved using technologies related to computer science. Features extracted from a face are processed
and compared with similarly processed faces present in the database. If a face is recognized it is
known or the system may show a similar face existing in database else it is unknown. In surveillance
system if a unknown face appears more than one time then it is stored in database for further
recognition. These steps are very useful in criminal identification. In general, face recognition
techniques can be divided into two groups based on the face representation they use appearance-based,
which uses holistic texture features and is applied to either whole-face or specific regions in a face image
and feature-based, which uses permanent facial features (mouth, eyes, brows, cheeks etc.), and geometric
relationships between them.

1.5 Video image detection:
We present Face detection and counting system based on digital image-processing techniques.
These images can be taken by digital cameras installed at the top of existing Feature features. By
using the proposed approach, it is possible to detect the number of Faces waiting on each side of the
intersection, hence, providing the necessary information for optimal Feature management. Results
achieved after testing this methodology on three real intersections are promising, attaining high
accuracy while counting several Faces at the same time. Hence, the system is equivalent to
installing multiple inductive loops in all the streets of the intersection, but with lower installation
and maintenance costs.

1.6 Microcontroller 89c51
The 89c51 is a microcontroller manufactured by the Atmel Corporation. It is considered as a
general purpose microcontroller. This is due to its low cost and its industry standard instruction set.

1.7 Description
The AT89C51 is a low-power, high-performance CMOS 8-bit microcomputer with 4K bytes of
Flash programmable and erasable read only memory (PEROM). The device is manufactured using
Atmel’s high-dens it’s nonvolatile memory technology and is compatible with the industry-standard
MCS-51 instruction set and pin out. The on-chip Flash allows the program memory to be
reprogrammed in-system or by a conventional nonvolatile memory programmer. By combining a
versatile 8-bit CPU with Flash on a monolithic chip, the Atmel AT89C51 is a powerful
microcomputer which provides a highly-flexible and cost-effective solution to many embedded
control applications.


1.8 Features
4K Bytes of In-System Reprogrammable Flash Memory



Fully Static Operation: 0 Hz to 24 MHz



Three-level Program Memory Lock



128 x 8-bit Internal RAM



32 Programmable I/O Lines



Two 16-bit Timer/Counters



Six Interrupt Source

11

1.9 Data Sheet 89c51

12

Chapter 2

Image Processing

The field of digital image processing refers to processing digital images by means of a digital
computer. There is no general agreement among authors regarding where image processing stops
and other related areas, such as image analysis and computer vision, start. Sometimes a
distinction is made by defining image processing as a discipline in which both the input and
output of a process are images. We believe this to be a limiting and somewhat artificial
boundary. For example, under this definition, even the trivial task of computing the average
intensity of an image (which yields a single number) would not be considered an image
processing operation. On the other hand, there are fields such as computer vision whose ultimate
goal is to use computers to emulate human vision, including learning and being able to make
inferences and take actions based on visual inputs. This area itself is a branch of artificial
intelligence (AI) whose objective is to emulate human intelligence. The field of AI is in its
earliest stages of infancy in terms of development, with progress having been much slower than
originally anticipated. The area of image analysis (also called image understanding) is in
between image processing and computer vision.

2.1 Background:
One of the first applications of digital images was in the newspaper industry, when pictures were
first sent by submarine cable between London and New York. Introduction of the Bartlane cable
picture transmission system in the early 1920s reduced the time required to transport a picture
across the Atlantic from more than a week to less than three hours. Specialized printing
equipment coded pictures for cable transmission and then reconstructed them at the receiving
end.
Some of the initial problems in improving the visual quality of these early digital pictures were
related to the selection of printing procedures and the distribution of intensity levels. The
printing method used to obtain was abandoned toward the end of 1921 in favor of a technique
based on photographic reproduction made from tapes perforated at the telegraph receiving
terminal.

2.2 Types of Processing:
There are no clear-cut boundaries in the continuum from image processing at one end to
computer vision at the other. However, one useful paradigm is to consider three types of
computerized processes in this continuum: low-, mid-, and high-level processes.
13

2.2.1 Low-level processing:

Low-level processing involves primitive operations such as image preprocessing to reduce noise,
contrast enhancement, and image sharpening. A low-level process is characterized by the fact
that both its inputs and outputs are images.
2.2.2 Mid-level processing:

Mid-level processing on images involves tasks such as segmentation (partitioning an image into
regions or objects), description of those objects to reduce them to a form suitable for computer
processing, and classification (recognition) of individual objects. A mid-level process is
characterized by the fact that its inputs generally are images, but its outputs are attributes
extracted from those images (e.g., edges, contours, and the identity of individual objects).
2.2.3 Higher-level processing:

Higher-level processing involves “making sense” of an ensemble of recognized objects, as in
image analysis, and, at the far end of the continuum, performing the cognitive functions normally
associated with vision.

2.3 Components of an Image Processing System:
As recently as the mid-1980s, numerous models of image processing systems being sold
throughout the world were rather substantial peripheral device that attached to equally substantial
host computers. Late in the 1980s and early in the 1990s, the market shifted to image processing
hardware in the form of single boards designed to be compatible with industry standard buses
and to fit into engineering workstation cabinets and personal computers. In addition to lowering
costs, this market shift also served as a catalyst for a significant number of new companies
specializing in the development of software written specifically for image processing.
Although large-scale image processing systems still are being sold for massive imaging
applications, such as processing of satellite images, the trend continues toward miniaturizing and
blending of general-purpose small computers with specialized image processing hardware.
Figure 2.1 shows the basic components comprising a typical general purpose system used for
digital image processing. The function of each component is discussed in the following paragraphs, starting with image sensing.
With reference to sensing, two elements are required to acquire digital images. The first is a
physical device that is sensitive to the energy radiated by the object we wish to image. The
14

second, called a digitizer, is a device for converting the output of the physical sensing device into
digital form. For instance, in a digital video camera, the sensors produce an electrical output
proportional to feature intensity.

Problem domain
Figure 2.1 Components of general purpose image processing system.

Specialized image processing hardware usually consists of the digitizer just mentioned, plus
hardware that performs other primitive operations, such as an arithmetic logic unit (ALU), that
performs arithmetic and logical operations in parallel on entire images. One example of how an
ALU is used is in averaging images as quickly as they are digitized, for the purpose of noise
15

reduction. This type of hardware sometimes is called a front-end subsystem, and its most
distinguishing characteristic is speed. In other words, this unit performs functions that require
fast data throughputs (e.g., digitizing and averaging video images at 30 frames/s) that the typical
main computer cannot handle.
The computer in an image processing system is a general-purpose computer and can range from
a PC to a supercomputer. In dedicated applications, some-times custom computers are used to
achieve a required level of performance, but our interest here is on general-purpose image
processing systems. In these systems, almost any well-equipped PC-type machine is suitable for
off-line image processing tasks.
Software for image processing consists of specialized modules that perform specific tasks. A
well-designed package also includes the capability for the user to write code that, as a minimum,
utilizes the specialized modules. More sophisticated software packages allow the integration of
those modules and general-purpose software commands from at least one computer language.

2.4 Applications:
Today, there is almost no area of technical endeavor that is not impacted in some way by digital
image processing. We can cover only a few of these applications in the context and space of the
current discussion. However, limited as it is, the material presented in this section will leave no
doubt in your mind regarding the breadth and importance of digital image processing.
In this section numerous areas of application, each of which routinely utilizes the
digital image processing techniques. Many of the images shown in this section are used later in
one or more of the examples given in the book. The areas of application of digital image
processing are so varied that some form of organization is desirable in attempting to capture the
breadth of this field.

One of the simplest ways to develop a basic understanding of the extent of image processing
applications is to categorize images according to their source.

The principal energy source for images in use today is the electromagnetic energy spectrum.
Other important sources of energy include acoustic, ultrasonic, and electronic. Synthetic images,
used for modeling and visualization, are generated by computer.
Images based on radiation from the EM spectrum are the most familiar, especially images in the
X-ray and visual bands of the spectrum. Electromagnetic waves can be conceptualized as
propagating sinusoidal waves of varying wavelengths, or they can be thought of as a stream of
mass less particles, each traveling in a wavelike pattern and moving at the speed of feature. Each
mass less particle contains a certain amount (or bundle) of energy. Each bundle of energy is
called a photon.

16

Angiography is another major application in an area called contrast enhancement radiography.
This procedure is used to obtain images of blood vessels.
Applications of ultraviolet “feature” is varied .They include lithography, industrial inspection,
microscopy, lasers, biological imaging, and astronomical observations. We illustrate imaging in
this band with examples from microscopy and astronomy. Ultraviolet feature is used in
fluorescence microscopy, one of the fastest growing areas of microscopy. Fluorescence is a
phenomenon discovered in the middle of the nineteenth century, when it was first observed that
the mineral fluorspar fluoresces when ultraviolet feature is directed upon it.
Considering that the visual band of the electromagnetic spectrum is the most familiar in all our
activities, it is not surprising that imaging in this band outweighs by far all the others in terms of
breadth of application. The infrared band often is used in conjunction with visual imaging, so we
have grouped the visible and infrared bands in this section for the purpose of illustration. We
consider in the following discussion applications in feature microscopy, astronomy, remote
sensing, industry, and law enforcement.
Image processing has been applied to Feature analysis in recent years, with different goals. In
this report, a new approach is presented for extracting vehicular speed information, given a
sequence of real-time Feature images.

17

CHAPTER 3

FACE DETECTION

3.1 Introduction
There is a lot of work done on the problem of Face detection. Several of the existing techniques
can robustly detect with extremely high accuracy. Due to various poses, different scales,
different expressions, featureing conditions, complex background and orientation, the detection
rate and real time of exist methods have not reached the satisfaction. For Face detection, we use
Viola-Jones method. Their method is the most common method for Face detection. In this
chapter we will explain the Face detection process.
There are four key concepts in Face detection.
3.2. Computing simple rectangular features, called Haar features.
3.3. An Integral image is first created which is then used for rapid computation of features.
3.4. The AdaBoost machine-learning method is used to select and train weak classifiers from the
feature set.
3.5. A cascaded classifier is constructed by combining many features (week classifiers)
efficiently.
Now we will explain these concepts in more detail and describe the Face detection procedure.

3.2 Rectangular features (Haar features):
Face detection procedure classifies images based on the value of simple features. Motivations for
18

using features rather than the pixels directly are that features can act to encode ad-hoc domain
knowledge that is difficult to learn using a finite quantity of training data. Also feature-based
system operates much faster than a pixel-based system. Most commonly used features are shown
in Fig 3.1. The value of a two-rectangle feature is the difference between the sums of the pixels
within two rectangular regions. A three- rectangle feature computes the sum within two outside
rectangles subtracted from the sum in a center rectangle. Finally a four-rectangle feature
computes the difference between diagonal pairs of rectangles. Given that the base resolution of
the detector is 24 × 24, the exhaustive set of rectangle features is quite large, 160,000.

Figure.3-1. Example rectangle features shown relative to the enclosing detection window. The sum of the
pixels which lie within the white rectangles are subtracted from the sum of pixels in the grey rectangles. Tworectangle features are shown in (A) and (B). Figure (C) shows a three-rectangle feature, and (D) a fourrectangle feature.

3.3 Feature Count:
A feature count is a count of feature along a particular Detection Window, either done
electronically or by people counting by the side of the Detection Window. Feature counts can be
used by local councils to identify which routes are used most, and to either improve that
Detection Window or provide an alternative if there is an excessive amount of Feature. Also,
some geographic fieldwork involves a Feature count. They are useful for comparing two or more
Detection Windows.

19

Figure 4.1 shows the counting of Faces on the Detection Window

3.4 An Integral Image for rapid feature computation:

Rectangle features can be computed very rapidly using integral image. The integral image at
location x, y is the sum of the pixels above and to the left of x, y, inclusive:

ii(x,y) =∑𝒙′≤𝒙,𝒚′≤𝒚 𝒊(𝒙′ , 𝒚′)

(3-1)

where ii(x,y) is the integral image and i(x,y) is the original image. See figure (3.2). Using the
following pair of recurrences:

s(x,y) = s(x,y-1) + i(x,y)
ii(x,y) = ii(x-1,y) + s(x,y)

(3-2)

In an integral image the value at pixel (x,y) is the sum of pixels above and to the left of (x,y).

20

Input image

Integral image

Figure 3-2. shows how an integral image formed from the input image

The integral image is computed in one pass over the original image. Once integral image is
computed, any rectangular sum can be computed in four array references (see Fig. 3.4). Clearly
the difference between two rectangular sums can be computed in eight references. Since the tworectangle features defined above involve adjacent rectangular sums they can be computed in six
array references, eight in the case of the three-rectangle features, and nine for four-rectangle
features.

Figure 3-3. The value of the integral image at point (x, y) is the sum
of all the pixels above and to the left.

21

Figure.3-4.The sum of the pixels within rectangle D can be computed with four array references. The value of
the integral image at location 1 is the sum of the pixels in rectangle A. The value at location 2 is A + B, at
location 3 is A + C, and at location 4 is A + B + C + D. The sum within D can be computed as 4 + 1 − (2 + 3).

By seeing above figures, now we understand the concept of Integral Image which is used in the
computation of rectangular features. Now we will discuss the AdaBoost machine-learning
algorithm which was first time used by Viola and Jones in 2004.

3.5 The AdaBoost Machine –learning method:
Viola and Jones had developed first real time detection system using AdaBoost in 2004.
AdaBoost is an algorithm for constructing a “strong” classifier as a linear combination of “weak”
classifiers. This algorithm is a method to improve the accuracy based on a series of weak
classifiers stage by stage.
Given a feature set and a training set of positive and negative images, a classification function
can be developed. Recall that there are 160,000 rectangle features associated with each image
sub-window, a number far larger than the number of pixels. A very small number of these
features can be combined to form an effective classifier. A variant of AdaBoost is used both to
select the features and to train the classifier. Basically AdaBoost learning algorithm is used to
boost the classification performance of a simple learning algorithm (e.g., it might be used to
boost the performance of a simple perceptron). It does this by combining a collection of weak
classification functions to form a stronger classifier. In the language of boosting the simple
learning algorithm is called a weak learner. In order for the weak learner to be boosted, it is
called upon to solve a sequence of learning problems. After the first round of learning, the
examples are re-weighted in order to highfeature those which were incorrectly classified by the
previous weak classifier. The weak learning algorithm is designed to select the single rectangle
feature which best separates the positive and negative examples. For each feature, the weak
22

learner determines the optimal threshold classification function, such that the minimum numbers
of examples are misclassified. A weak classifier (h(x, f, p, θ)) thus consists of a feature (f), a
threshold (θ) and a polarity (p) indicating the direction of the inequality:

( )
( , , , )

{

}

(3-3)

Here x is a 24 × 24 pixel sub-window of an image. In practice no single feature can perform the
classification task with low error. Table 3.1 shows the learning algorithm.

3.5.1 Working of Adaboost Algorithm:

AdaBoost starts with a uniform distribution of “weights” over training examples.
Select the classifier with the lowest weighted error (i.e. a “weak” classifier)
Increase the weights on the training examples that were misclassified.
Repeat the procedure.
At the end, personefully make a linear combination of the weak classifiers obtained at all
iterations

Table 3.1. The boosting algorithm for learning a query online. T hypotheses are constructed each using a
single feature. The final hypothesis is a weighted linear combination of the T hypotheses where the weights
are inversely proportional to the training errors.



Given example images (x1,y1),……… (xn,yn) where yi = 0, 1 for negative and positive
examples respectively.



Initialize weights w1,I =



negatives and positives respectively.
For t = 1,………….T.
1. Normalize the weights

,

for yi = 0, 1 respectively where m and l are the number of

So, that wt is the probability distribution.
2. For each feature j, train a classifier hj which is restricted to using a single feature. The
error is evaluated with respect to wt,

23

3. Choose the classifier ht with lowest error €i.
4. Update the weights:
where ei = 0 if example xi is classifier correctly, ei = 1 otherwise, and

5. The final strong classifier is:

Since final classifier is a weighted combination of weak classifiers. Pictorial representation of
AdaBoost classification phenomenon is shown in Figure 3-4. The AdaBoost learning algorithm
initially maintains a uniform distribution of weights over each training samples. In the first
iteration, the algorithm trains a weak classifier using one Haar-like feature that achieves the best
recognition performance for the training samples. In the second iteration, the training samples
that were misclassified by the first weak classifier receive higher weights so that the newly
selected Haar-like feature must focus more computation power towards these misclassified
samples. The iteration goes on and the final result is a cascade of linear combinations of the
selected weak classifiers, i.e. a strong classifier, which achieves the required accuracy.

3.6 The Cascaded Classifier:
In practical implementation, the Attention cascade is employed to speed up the performance of
the learning algorithm. In the first stage of the training process, the threshold of the weak
classifier is adjusted low enough so that 100% of the target objects can be detected while keeping
the false negative rate close to zero. The trade-off of a low threshold is that a higher false
positive detection rate will accompany. A cascade of classifiers achieves increased detection
performance while reducing computation time. Smaller and efficient boosted classifiers can be
constructed which reject many of the negative sub-windows while detecting almost all positive
instances. Simpler classifiers are used to reject the most of sub-windows before more complex
classifiers used to achieve low false positive rates. Stages in the cascade are constructed by
training classifiers using AdaBoost. Starting with a two-feature strong classifier, an effective
filter can be obtained by adjusting the strong classifier threshold to minimize false negatives. The
initial AdaBoost threshold is designed to yield a low error rate on the training data. A lower
threshold yields higher detection rates and higher false positive rates. Based on performance
measured using a validation training set, the two-feature classifier can be adjusted to detect
24

100% of the players/ s with a false positive rate of 50%. Fig. 3.5 shows a description of the
features used in this classifier.

Figure. 3.6 This feature is selected by AdaBoost. The feature measures the difference in intensity between the
region below headfeatures and a region of shadow between the tyres. The feature capitalizes on the
observation that the headfeature region is often featureer than the shadow region.

Training a cascade:

To design a cascade we must choose:
 Number of stages in cascade (strong classifiers).
 Number of features of each strong classifier.
25




Threshold of each strong classifier Optimization problem:
Can we find optimum combination?

3.6.2 Optimum Combination:

Since finding optimum combination is extremely difficult. Viola & Jones suggested a heuristic
algorithm for the cascade training.
Manual Tweaking:
 select fi (Maximum Acceptable False Positive rate / stage)
 select di (Minimum Acceptable True Positive rate / stage)
 select Ftarget (Target Overall False Positive rate)
Until Ftarget is met:
 Add new stage:
 Until fi, di rates are met for this stage
 Keep adding features & train new strong classifier with AdaBoost.
Table 3-2 shows the shows pseudo code for building a cascade detector.








User selects values for f, the maximum acceptable false positive rate per layer and d, the
minimum acceptable detection rate per layer.
User selects target overall false positive rate[1], Ftarget.
P = set of positive examples
N = set of negative examples
F0 = 1.0; D0 = 1.0.
I=0



While Fi > Ftarget





i← i + 1
ni = 0; Fi = Fi -1
while Fi > f x Fi -1







ni ← ni + 1
Use P and N to train a classifier with ni features using AdaBoost.
Evaluate current cascaded classifier on validation set to determine Fi and Di.
Decrease threshold for the ith classifier until the current cascaded classifier has
a detection rate of at least d x Di -1 (this also affects Fi)
N←Ø
If Fi > Ftarget then evaluate the current cascaded detector on the set of nonFace images and put any false detections into the set N

26

The overall form of the detection process is that of a degenerate decision tree, what is called a
“cascade” (Quinlan, 1986) (see Fig. 2.6). A positive result from the first classifier starts a second
classifier which has also been adjusted to achieve very high detection rates. A positive result
from the second classifier triggers evaluation of a third classifier, and so on. A negative outcome
at any point means an immediate rejection of the sub-window. The structure of the cascade
utilized the fact that within any single image an overwhelming majority of sub-windows are
negative. So the cascade tries to reject as many negatives as possible at the earliest stage
possible. We developed 20 stage classifier and training took 30 hours to complete on Pentium IV
system with 3 GHz processor and 2 GB RAM.

Figure. 3.6. Schematic depiction of a detection cascade. A series of classifiers are applied to every subwindow. The initial classifier eliminates a large number of negative examples with very little processing.
Subsequent layers eliminate additional negatives but require additional computation. After several stages of
processing the numbers of sub-windows have been reduced radically. Further processing can take any form
such as additional stages of the cascade (as in our detection system) or an alternative detection system.

While a positive instance will starts the evaluation of every classifier in the cascade, this is a very
rare event. Like a decision tree, subsequent classifiers are trained using those examples which
pass through all the previous stages. So the second classifier performs a more difficult task than
the first. The examples which passes through the first stage are difficult than typical examples.
We collected 1000 positive samples of Faces and conditions for training with size of 24x24. To
obtain a robust classifier we used good quality positive samples for training. Figure 3.8 shows
some of the positive samples used in our work. Similarly 1500 negative images were used of size
640x480. Figure 3.9 shows some of the negative images used in our training process

27

Figure. 3.7. Positive Samples used in training process

Figure. 3.8. Negative Samples used in training process

Now the training algorithm based on AdaBoost learning algorithm takes a set of “Positive” and
“Negative” samples and generates a classifier and detect the Face.

28

Chapter 6

Conclusion

In this work we developed an application for high ways. Our application can vigorously detect
Faces, count them and measure their speed. We achieved precise and robust Face detection,
count and speed measurement. For Face detection and count we used Haar-like features and
AdaBoost algorithm for both feature selection and classification. The Face speed measurement
system used AdaBoost algorithm with optical flow method. For Face recognition general
databases have been used. Tested Face images videos include different featureing conditions. In
our application we have tested the Face video of different image sizes. The
recognition system efficiently worked on most difficult scenario of sample size of 5 x 5. The
results have shown that the developed application is fully applicable in
real world environments.

6.1 DETECTION:

Figure 6.1 shows Face detection

It is the result of detection. It only detects the respective Face and does not detect any other thing
in the surrounding.

6.2 SPEED MEASUREMENT
It is Face speed measurement. It is done by distance time formula i.e.,
speed=(distance2-distance1)/time.
1) Distance1 and distance 2 2) Time known.
29

REFRENCES:
1)Detection Window Feature Data: Collection Methods and Applications Guillaume Leduc
Working Papers on Energy, Transport and Climate Change

2) Automatic player detection and recognition in images using adaboost Zahid Mahmood
August 2011.
3) Digital Image Processing Third Edition Rafael C. Gonzales published in 7th October 2007.
4) Robust Real-time Object Detection Paul Viola Michael Jones 13 July 2001

Website:
5) http://ntl.bts.gov/DOCS/arizona_report.html

6) http://en.wikipedia.org/wiki/OpenCV

7) http://en.wikipedia.org/wiki/Optical_flow

30

31

Appendix A

OpenCV (Open Source Computer Vision Library) is a library of programming functions mainly
aimed at real-time computer vision, developed by Intel, and now supported by Willow
Garage and Itseez. It is free for use under the open source BSD license. The library is crossplatform. It focuses mainly on real-time image processing. If the library finds Intel's Integrated
Performance Primitives on the system, it will use these proprietary optimized routines to
accelerate itself.
Officially launched in 1999, the OpenCV project was initially an Intel Research initiative to
advance CPU-intensive applications, part of a series of projects including real-time ray
tracing and 3D display walls. The main contributors to the project included a number of
optimization experts in Intel Russia, as well as Intel’s Performance Library Team. In the early
days of OpenCV, the goals of the project were described as


Advance vision research by providing not only open but also optimized code for basic vision
infrastructure. No more reinventing the wheel.



Disseminate vision knowledge by providing a common infrastructure that developers could
build on, so that code would be more readily readable and transferable.



Advance vision-based commercial applications by making portable, performance-optimized
code available for free—with a license that did not require to be open or free themselves.

The first alpha version of OpenCV was released to the public at the IEEE Conference on
Computer Vision and Pattern Recognition in 2000, and five betas were released between 2001
and 2005. The first 1.0 version was released in 2006. In mid-2008, OpenCV obtained corporate
support from Willow Garage, and is now again under active development. A version 1.1 "prerelease" was released in October 2008.
The second major release of the OpenCV was on October 2009. OpenCV 2 includes major
changes to the C++ inter, aiming at easier, more type-safe patterns, new functions, and better
implementations for existing ones in terms of performance (especially on multi-core systems).
Official releases now occur every six months and development is now done by an independent
Russian team supported by commercial corporations.
In August 2012, support for OpenCV was taken over by a non-profit foundation, OpenCV.org,
which maintains a developer and user site.

32

APPLICATIONS:
OpenCV's applications areas include:













2D and 3D feature toolkits
Egomotion estimation
Facial recognition system
Gesture recognition
Mobile robotics
Motion understanding
Object identification
Segmentation and Recognition
Stereopsis Stereo vision: depth perception from 2 cameras
Structure from motion (SFM)
Motion tracking

OS support:
OpenCV runs on Windows, Android, Maemo, FreeBSD, OpenBSD, BlackBerry ,Linux and OS
X. The user can get official releases from SourceForge, or take the current snapshot
under SVN from there. OpenCV uses CMake.

Description:

The main objective for all computer vision processes

Syntax:

OpenCV(parent);

Fields:
BILATERAL

Blur method

BLUR

Blur method

BUFFER

Type of Image

CASCADE_FRONTAL_ALT

Standard Haar classifier cascade file used for
object detection
33

CASCADE_FRONTAL_ALT2

Standard Haar classifier cascade file used for
object detection

CASCADE_FRONTAL_ALT_TREE

Standard Haar classifier cascade file used for
object detection

CASCADE_FRONTAL_DEFAULT

Standard Haar classifier cascade file used for
object detection

CASCADE_FULLBODY

Standard Haar classifier cascade file used for
object detection

Standa

objec
CASCADE_UPPERBODY

Standard Haar classifier cascade file used for
object detection

FLIP_BOTH

Flip mode

FLIP_HORIZONTAL

Flip mode

FLIP_VERTICAL

Flip mode

GAUSSIAN

Blur method

GRAY

Color space of image

HAAR_DO_CANNY_PRUNING

Haar classifier flag

HAAR_DO_ROUGH_SEARCH

Haar classifier flag

HAAR_FIND_BIGGEST_OBJECT

Haar classifier flag

HAAR_SCALE_IMAGE

Haar classifier flag

INTER_AREA

Interpolation method

34

INTER_CUBIC

Interpolation method

INTER_LINEAR

Interpolation method

INTER_NN

Interpolation method

MAX_VERTICES

The maximum number of contour points available
to blob detection (by default)

MEDIAN

Blur method

MEMORY

Type of Image

MOVIE_FRAMES

Movie info selector (not yet implemented)

MOVIE_MILLISECONDS

Movie info selector (not yet implemented)

MOVIE_RATIO

Movie info selector (not yet implemented)

RGB

Color space of image

SOURCE

Type of Image

THRESH_BINARY

Thresholding method

THRESH_BINARY_INV

Thresholding method

THRESH_OTSU

Thresholding method

THRESH_TOZERO

Thresholding method

THRESH_TOZERO_INV

Thresholding method

THRESH_TRUNC

Thresholding method

height

OpenCV image/buffer height

width

OpenCV image/buffer width

35

ROI()

Set image region of interest to the given rectangle.

absDiff()

Calculate the absolute difference between the image in
memory and the current image.

allocate()

Allocate required buffer with the given size.

blobs()

Blob and contour detection.

blur()

Smooth the image in one of several ways.

brightness()

Adjust the image brightness with the specified value
(in range of -128 to 128).

capture()

Allocate and initialize resources for reading a video
stream from a camera.

cascade()

Load into memory the descriptor file for a trained cascade classifier.

contrast()

Adjust the image contrast with the specified value
(in range of -128 to 128).

convert()

Convert the current image from one colorspace to another.

copy()

Copy the image (or a part of it) into the current OpenCV
buffer (or a part of it).

detect()

Detect object(s) in the current image depending on the
current cascade description.

flip()

Flip the current image around vertical, horizontal or both axes.

image()

Return the current (or specified) OpenCV image

interpolation()

Set global interpolation method.

invert()

Invert image.

36

jump()

Jump to a specified movie frame.

loadImage()

Load an image from the specified file.

movie()

Allocate and initialize resources for reading a video
file from the specified file name.

pixels()

Retrieve cuurent (or specified) image data.

read()

Grab a new frame from the input camera or a movie file.

remember()

Place the image (original or current) in memory.

restore()

Revert to the original image.

stop()

Stop OpenCV process.

threshold()

Apply fixed-level threshold to the current image.

37

Appendix B

C++ language:
C++ (pronounced "see plus plus") is a statically typed, free-form, multi-paradigm, compiled,
general-purpose programming language. It is regarded as an intermediate-level language, as it
comprises
both high-level and low-level language
features. Developed
by
Bjarne
Stroustrup starting in 1979 at Bell Labs, C++ was originally named C with Classes,
adding object oriented features, such as classes, and other enhancements to the C programming
language. The language was renamed C++ in 1983,as a pun involving the increment operator.
C++ is one of the most popular programming languages and is implemented on a wide variety of
hardware and operating system platforms. As an efficient compiler to native code, its application
domains include systems software, application software, device drivers, embedded software,
high-performance server and client applications, and entertainment software such as video
games.Several groups provide both free and proprietary C++compiler software, including
the GNU Project, LLVM,Microsoft, Intel and Embarcadero Technologies. C++ has greatly
influenced many other popular programming languages, most notably C# and Java. Other
successful languages such as Objective-C use a very different syntax and approach to adding
classes to C.
C++ is also used for hardware design, where the design is initially described in C++, then
analyzed, architecturally constrained, and scheduled to create a register-transfer level hardware
description language via high-level synthesis.
The language began as enhancements to C, first adding classes, then virtual functions, operator
overloading, multiple inheritance, templates and exception handling, among other features. After
years of development, the C++ programming language standard was ratified in 1998 as ISO/IEC
14882:1998.

Philosophy:
In The Design and Evolution of C++ (1994), Bjarne Stroustrup describes some rules that he used
for the design of C++.


C++ is designed to be a statically typed, general-purpose language that is as efficient and
portable as C

38









C++ is designed to directly and comprehensively support multiple programming styles
(procedural programming, data abstraction, object-oriented programming, and generic
programming)
C++ is designed to give the programmer choice, even if this makes it possible for the
programmer to choose incorrectly
C++ is designed to be compatible with C as much as possible, therefore providing a smooth
transition from C
C++ avoids features that are platform specific or not general purpose
C++ does not incur overhead for features that are not used (the "zero-overhead principle")
C++ is designed to function without a sophisticated programming environment

Inside the C++ Object Model (Lippman, 1996) describes how compilers may convert C++
program statements into an in-memory layout. Compiler authors are, however, free to implement
the standard in their own manner.

Operators and operator overloading
Operators that cannot be overloaded

Operators

Symbols

Scope Resolution Operator

::

Conditional Operator

:?

Dot Operator

.

Member Selection Operator

.*

“sizeof” Operator

Sizeof

“typeid” Operator

typeid

39

C++ provides more than 35 operators, covering basic arithmetic, bit manipulation, indirection,
comparisons, logical operations and others. Almost all operators can be overloaded for userdefined types, with a few notable exceptions such as member access (. and .*) as well as the
conditional operator. The rich set of overloadable operators is central to using user created types
in C++ as well and as easily as built in types (so that the user using them cannot tell the
difference). The overloadable operators are also an essential part of many advanced C++
programming techniques, such as smart pointers. Overloading an operator does not change the
precedence of calculations involving the operator, nor does it change the number of operands
that the operator uses (any operand may however be ignored by the operator, though it will be
evaluated prior to execution). Overloaded "&&" and "||" operators lose their short-circuit
evaluation property.

40

APPENDIX C
FALSE POSITIVE RATE:
In statistics when performing multiple comparisons, the term false positive ratio, also known as
false alarm ratio, usually refers to the probability of falsely rejecting the null hypothesis for a
particular test. The false positive rate usually refers to the expectancy of the positive ratio.

41

REFRENCES:
1)Detection Window Feature Data: Collection Methods and Applications Guillaume Leduc
Working Papers on Energy, Transport and Climate Change

2) Automatic player detection and recognition in images using adaboost Zahid Mahmood
August 2011.
3) Digital Image Processing Third Edition Rafael C. Gonzales published in 7th October 2007.
4) Robust Real-time Object Detection Paul Viola Michael Jones 13 July 2001

Website:
5) http://ntl.bts.gov/DOCS/arizona_report.html

6) http://en.wikipedia.org/wiki/OpenCV

7) http://en.wikipedia.org/wiki/Optical_flow

42

43

44