GOVERNMENT ENGINEERING COLLEGE,

GANDHINAGR

TOPIC: REAL TIME ANTI-THEFT VEHICLE

SYSTEM USING ARM 7

PRESENTED BY: -

NAME: DISHANT PAL

ENROLLMENT NO.: 220130111029
BATCH: EC A1
 Anti-Theft Vehicle System using
ARM 7

Abstract
The development of vehicular design brings public many conveniences in life but also brings many
problems at the same time, for example, traffic congestion, difficult in monitoring dispersive vehicle,
theft and other series of problems [3]. Due to the insecure environment the ratio of vehicle theft increases
rapidly. Because of this is manufacturers of luxury automobiles has the responsibilities for taking steps to
ensure the authorization for the owners and also in built the anti-theft system to prevent the vehicle from
theft. The proposed security system for smart cars used to prevent them from loss or theft using
Advanced RISC Machine (ARM) processor. It performs the real time user authentication (driver, who
starts the car engine) using face recognition, using the Principle Component Analysis (PCA) algorithm.
According to the comparison result (authentic or not), ARM processor triggers certain actions. If the
result is not authentic means ARM produces the signal to block the car access.

Keywords: PCA, FDS, NOOBS

1. Introduction
Automatic face recognition is a widely used biological recognition technology. In comparison
with other identification methods, face recognition has direct, Friendly and convenient
features. Human can visually identify people by human face. People can be fairly identified
even in the very serious visual stimulated situation [2]. Now the market share of face
recognition is just less than the fingerprint recognition and the proportion are increasing,
which have broken the situation that fingerprints recognition monopolized the market in the
international biological recognition market. Therefore, it is very important for the research of
face recognition technology. This project, which is based on ARM11 and Linux operating
system, is a portable device and meets the requirements of up to date Identity Authentication;
it also can be used in many areas. The embedded face recognition system is based on ARM 11
development board, including transplantation of Linux operating system, the development of
drivers, detecting face by using face class Haar feature, and then recognizing face by using
PCA transform algorithm [1].

Block Diagram of Proposed System

A. Overview of the project
In this project, we propose an extendable emergency response system for smart car to prevent
them from loss ortheft as shown in fig. 1 using Advanced RISC Machine (ARM) processor
Correspondence: (RISC means Reduced Instruction Set Computing). In this method, the Face Detection
Dishant Pal Subsystem (FDS) aims at detect somebody's face (who try to access the car).By using PCA
Government College of algorithm we can get the common eigen values of the person and it compares the image by
Engineering finding the nearest value in some mathematical form which as like a function. If the person
Gandhinagar matches vehicle starts [5].

Fig 1: Block Diagram of Proposed system

B. Face Detection System (FDS) D. Embedded Control System
Face recognition is a non-intrusive method, and facial The ARM is a 32-bit Reduced Instruction Set Computer
attributes are probably the most common biometric features (RISC) Instruction Set Architecture (ISA) developed by
used by humans to recognize one another. The applications ARM Holdings. It was known as the Advanced RISC
of facial recognition range from a static, controlled Machine, and before that as the Acorn RISC Machine. The
authentication to a dynamic, uncontrolled face identification relative simplicity of ARM processors made them suitable
in a cluttered background. While the authentication for low power applications. This has made them dominant in
performance of the face recognition systems that are the mobile and embedded electronics market as relatively
commercially available is reasonable, they impose a number low cost and small microprocessors and microcontrollers
of restrictions on how the facial images are obtained, often
requiring a fixed and simple background with controlled IMPLEMENTATION OF SYSTEM
illumination. These system also have difficulty in matching System is based on ARM7 LPC2148 (Raspberry Pi Model
face images captured from two different views, under B+ board), which is a low cost, credit-card sized computer.
different illumination conditions, and at different times. 'th' Based on this hardware platform, Embedded Linux
is the threshold value depending on the real time operating system and drivers are developed firstly, and then
environment (varies from 300- 400). lt is questionable face recognition system is achieved on the operating system.
whether the face itself, without any contextual information,
is a sufficient basis for recognizing a person from a large A. Transplantation of Embedded Linux
number of identities with an extremely high level of New Out Of the Box Software (NOOBS) is an easy
confidence. operating system install manager for the Raspberry Pi.
NOOBS is designed to make it easy to select and install
C. Principle Component Analysis (PCA) operating systems for the Raspberry Pi without having
The purpose of PCA is to reduce the large dimensionality of worry about manually imaging your SD card. On first boot
the data space (observed variables) to the smaller intrinsic NOOBS will format your SD card & allow install Raspbian.
dimensionality of feature space (independent variables), Raspbian is free operating system based on Debian
which are needed to describe the data economically. The optimized for the Raspberry Pi hardware. An operating
main idea of using PCA for face recognition is to express the system is the set of basic programs & utilities that make
large 10 vector of pixels constructed from 15 facial image your Raspberry Pi run.
into the compact principal components of the feature space.
This can be called Eigen face Projection. B. Embedded Human Face Recognition System

Fig 2: Flow chart of face recognition system

In the above Figure 2 the first step is to detect human face. Classification classifier of "first weight and then light". At
Face detection is one of the main technical parts of the first, using the simple strong classifier which includes more
automatic face recognition system. Face detection needs to important features to find non-face windows and delete
determine whether the given image is a face, if it is them. As the importance of features declines, the number of
someone’s face, and then return the position of the face. In classifiers is increasing, but the number of detected windows
order to better recognize all of the faces in the image, the is declining.
system use face class Haar feature to detect human face [1].
After having gotten all of the faces, the next step is finding B. Human Face Image Recognition Based on PCA
the special face. Compared with traditional face recognition If the direct use of the above high dimensional Gabor feature
system, in this thesis the embedded face recognition system vector for classifier training and image recognition, will
is based on the Gabor feature extraction algorithms of the cause the disaster of high dimensions. Therefore, we need to
face region and principal component analysis (PCA) with reduce the dimensions of high dimensional Gabor feature
nearest neighbour classifier as the core algorithms of the vector. PCA is a statistical dimensionality reduction method,
system identification. Face samples are pre-processed for which produces the optimal linear least squares
extracting the Gabor feature vector, by using PCA to reduce decomposing of a training set. Kirby and Sirovich (1990)
the feature vector’s dimension and we can get face PCA applied PCA to representing faces and Turk and Pentland
subspace of the sample Gabor feature, we call it face space. (1991) extended PCA to recognizing faces. The algorithm is
And then the input faces are projected into the subspace of based on an information theory approach that decomposes
face PCA and get image face. The distance from face space face images into a small set of characteristic feature images
to image faces as a measure of face similarity. called “eigenfaces”, which may be thought of as the
principal components of the initial training set of face
Analysis of the Main Algorithms images. Automatically learning and later recognizing new
A. Face Detection Base on Haar Feature faces is practical within this framework. Recognition under
Viola [1] wrote a paper which has been published in 2001, widely varying conditions is achieved by training on a
can be said as watershed of real time face detection limited number of characteristic views. The approach has
technology, they implement the real time face detection by advantages over other face recognition schemes in its speed
integrating Adaboost and cascade algorithms, face detection and simplicity, learning capacity, and insensitivity to small
is beginning to make practical. Papageorgiou and viola or gradual changes in the face image. The main idea of the
extract features from images based on the application of PCA (or Karhunen-Loeve expansion) is to find the vectors
wavelet transform, and put forth the original concept of local that best account for the distribution of face images within
Haar like features. The feature database contains 3 types and the entire image space. These vectors define the subspace of
4 different forms of features. The 3 types: 2 rectangles face images, which we call “face space”. Each vector is of
feature, 3 rectangles feature, and 4 rectangles feature. The 4 length , describes a by image. Because these vectors are the
forms are shown in Figure 3. eigenvectors of the covariance matrix corresponding to the
original face images, and because they are face like in
appearance, we refer to them as “eigenfaces”.
Use

System test
Fig 3: The original forms of Haar-like features A. Face Detection
The face class Haar feature is used to detect human faces. As
Now we generate classifiers which are based on the basic shown in fig. 4
principles of Adaboost face detection.
Simple classifier: The input window x, is an eigenvalue of
the jth rectangle feature, the jth feature is generated into the
form of a simple classifier: indicated the value of a simple
classifier; is the threshold; indicated the direction of
inequality, the value is 1.
Strong classifier: the learning process of Adaboost algorithm
can be understood as “the process of greedy feature
selection." For solving a problem by weighted voting
mechanisms, with plenty of the weighted combination of
classification function to make the determination. The key of
algorithm is to give the good classification function a bigger
weight value; the bad one is given a smaller weight value.
Adaboost algorithm is a very good way that can find out a
few features to effectively classify target.
Cascade strong classifier: By Adaboost algorithm the strong Fig 4: original images
classifier is made of the important features, and can be used
to detect face. During the testing process due to have to Scan Final images of face detection by using face class Haar
all different sizes windows in tested image’s every position, feature are shown in fig. 5
the number of detection windows are so many. During the
process of actual facedetection we should adopt
B. Training Data of PCA
Now projecting the sample images into subspace,
gettingThe training data of PCA as shown in fig. 6
Then real-time getting human face image, using Euclidean
Distance measures the human face image, and testing the

Fig 7: The sample images & training data of PCA

100%

80%

60% Th3

Th2
40%
Fig 5: final images of face detection Th1
20%

0%
F1 F2 F3 F4 F5

Fig 8: The sample images & training data of PCA

Code implemented

CAR_ANTITHEFT_SYSTEM, CODE, READONLY

ENTRY ; Marks the entry point of the code

; Pin Configuration (Assuming P0.4 for sensor, P0.5 for alarm,

and P0.6 for relay)
Fig 6: The sample images & training data of PCA SENSOR_PIN EQU 0x00000010 ; P0.4 for motion sensor
(input)
Similarity of the training faces and the image. If the ALARM_PIN EQU 0x00000020 ; P0.5 for alarm (output)
measured difference is in the range of threshold value, it RELAY_PIN EQU 0x00000040 ; P0.6 for relay (output)
indicates real-time image meets the requirements and
the image exists in the face space, and then finding the ; Set up the GPIO pin directions
most similar face in the face space and outputting the LDR R0, =0xE0028000 ; Address of GPIO0 IODIR register
corresponding personal information, or the image can’t (for direction setup)
be recognized and system logs out. The number of LDR R1, [R0] ; Read current value of IODIR ORR
training images can be bigger if it is needed for R1, R1, ALARM_PIN ; Set P0.5 as output (alarm)ORR R1,
expanding the scope of recognition. The several system R1, RELAY_PIN ; Set P0.6 as output (relay) BIC R1, R1,
tests steps show that the system can recognize face to SENSOR_PIN ; Set P0.4 as input (sensor)
meet the requirements and get the corresponding STR R1, [R0] ; Store updated value to GPIO0 IODIR
person’s information.
; Main loop
Hardware Setup LOOP:
As shown in figure 7 Hardware set up consist ARM 7 LDR R0, =0xE0028004 ; Address of GPIO0 PIN register (to
Raspberry Pi Board with Raspbian Camera module and read sensor input)
motor will start and stop with respect to Known and LDR R1, [R0] ; Read the status of GPIO0 pins
Unknown person. Graph shown in figure 8 which is
Training data set vs. recognition rate. Implemented TST R1, SENSOR_PIN ; Test if the sensor (P0.4)
System provides 70% to 90% recognition accuracy. detectsmotion
BEQ NO_THEFT ; If no motion detected, branch to
NO_THEFT Conclusion

; If theft detected, activate alarm and disable ignition via relay The present method the camera captures owner’s image
LDR R0, =0xE0028008 ; GPIO0 SET register address only. If the owner’s relatives or friends want to start the
LDR R1, [R0] vehicle it will not start. To overcome this one, we can
ORR R1, R1, ALARM_PIN ; Turn on alarm (P0.5) extend this project by storing multiple faces into the
ORR R1, R1, RELAY_PIN ; Turn on relay (P0.6, disablingignition) memory. If anyperson wants to start the vehicle, the
STR R1, [R0] camera compares the person’s image with the all-stored
images. If the result is matched the motor will start. In
; Delay to allow the alarm to ring for a while future we can extend this by sending the unknown
BL DELAY person’s image to the owner’s mobile as well as the
information to police control room for taking immediate
NO_THEFT: action.
; If no theft detected, turn off alarm and enable ignition LDR
R0, =0xE002800C ; GPIO0 CLR register addressLDR R1, References
[R0] 1. Shimin Wang, Jihua Ye College of Computer
BIC R1, R1, ALARM_PIN ; Turn off alarm (P0.5) Information Engineering “Research and
BIC R1, R1, RELAY_PIN ; Turn off relay (P0.6, enabling Implementationof Embedded Face Recognition
ignition) System Based on ARM 9” ©2010 IEEE W.-K. Chen,
STR R1, [R0] Linear Networks and Systems (Book style). Belmont,
CA: Wadsworth, 1993, pp. 123–135.
B LOOP ; Repeat the loop 2. Chellappa R,Wilson C L,Sirohey S “Human and
machine recognition of faces:A survey[J]”. Digital
DELAY: Object Identifier,1995,83(5):705-741.
MOV R2, #0xFFFFF ; Arbitrary delay count
DELAY_LOOP: 3. D. Narendra Singh (M.Tech,Ph.d), K. Tejaswi
SUBS R2, R2, #1 ; Decrement the delay count (M.Tech,) Dept. of Electronics Engineering Anurag
BNE DELAY_LOOP ; If not zero, continue delay group of Institutions, Hydrabad A.P. India Real
BX LR ; Return from subroutine Time

END 4. D. Narendra Singh (M.Tech,Ph.d), K. Tejaswi

(M.Tech,) Dept. of Electronics Engineering Anurag
group of Institutions, Hydrabad A.P. India Real
Time