Sie sind auf Seite 1von 6

International Journal of Communications and Engineering Volume 01 No.

1, Issue: 03 March2012

INTELLIGENT VEHICLE THEFT PREVENTION AND LOCATION FINDING SYSTEM USING FACE DETECTION BASED ON IMAGE PROCESSING
Miss K.Sangeetha M.E
Assistant Professor/ECE Surya Engineering College,

S.Arulraj1, G.Aravind2, M.M.Vignesh3


ECE, Final Year Surya Engineering College,

ABSTRACT:

The evident fact in the world is that the number of vehicles particularly cars and other automobiles escalating in terms of invention as well as theft attempts. Even though a lot of global and other manufacturing industry did lot of efforts to invent vehicle security system particularly for cars, the product doesnt reach greater heights because the car theft still increasing in numbers. Thus the more powerful security system is essential to prevent the vehicles from much smarter thieves. This project proposes an intelligent vehicle theft prevention and location finding system based on image processing to solve this issue. This system takes the photo of the driver and compares the captured driver face with database to check whether the driver authenticated not. If the driver is not an authenticated driver the captured image of the driver is send to the owners mobile through MMS. Then the owner can stop the vehicle by sending a SMS contains secret command to the GSM modem which placed in the vehicle. Thus the commands cease ignition system of the vehicle and made to stop the vehicle slowly. Then the vehicle location is send to the owners mobile through GPS. The timing and results are apt to owner and police to take proper action against the thief.
Keywords- Face detection, Face recognition, image processing, Global Positioning System (GPS), Global System for Mobile communication (GSM)

1 INTRODUCTION Access control security system plays a vital role in this contemporary era. Because it is must for the people to protect their belonging. The access control system requires some identification to access the users own database or the resources. Smartcards, tokens, encrypted keys, passwords and biometric systems are some of the different types of access control security system. Main disadvantage of these kind of access control security system is that the lacking of information about the user who is accessing the system because they depend only on data given to system. Biometric systems are most modern and sophisticated access control security systems. It is almost impossible to make a replica of distinctive human characteristics. There are different biometric technologies like voice recognition, signature recognition, retinal scanning, iris recognition, fingerprint and face recognition. Compared to all other biometric techniques advantages of face recognition include: Easy to install, also already existed image capturing devices can be used.

Page 1

International Journal of Communications and Engineering Volume 01 No.1, Issue: 03 March2012

It doesnt require any physical contact and also user can be identified easily. It is easier to obtain a photo rather than a finger print or iris scan. Face recognition and face detection system in access control system plays a major role in this system and it is discussed in detail. The system recognize driver entering the car and check whether the driver authenticated or not. A camera which installed in the takes the photo of the driver for the face recognition and detection system. For face detection morphological based approach is used. The database contains only the face portion of all drivers. Once database is created, these faces are passed into face recognition stage. As photos are taken in a constantly changing background, this may affect the accuracy of face detection algorithm. To avoid this problem, a template matching approach is used. In template matching, a template of size of frame is developed, with centre of frame where face is expected to be present alone is white and all other regions are black. For face recognition Eigen face approach is used. The system compares the face of driver entered, with faces in the database, which is computed from face detection stage. For face recognition Principal component analysis (PCA) algorithm is used. Principal component analysis (PCA) involves a mathematical procedure that transforms a number of possibly correlated variables into a number of uncorrelated variables called principal components, related to the original variables by an orthogonal transformation. PCA algorithm is used because it is one of the best classifying algorithm and it is simple to code PCA algorithm involves dimensionality reduction. Reduction in dimension demands less processing time and less memory. PCA algorithm classifies

a face as one belonging to database or not based on a threshold value. In this paper a different approach is used to determine the threshold which is explained in the following sections.

2 FACE DETECTION Face detection is the first step towards face recognition. Face detection helps to extract face region alone from the whole image. There are different algorithms present to detect face. Face can be detected using skin color. Main disadvantage of this approach is that region with color similar to skin will be wrongly determined as face. Neural networks can also be used for face detection. Here for face detection we use color segmentation and morphological process based algorithm. 2.1 Algorithm for face detection Face detection in general terms is defined as to isolate human faces from their background and exactly locate their position in an image. A webcam is used to take photo of the driver. At first fifteen photos of the driver with different pose will be taken. Then the photos which are taken in different poses are stored in the database for the face detection. Thus the face detection plays an important role and its algorithm and steps involved are explained in the following. These photos are treated with algorithm to extract face portion alone. 2.2 Face detection steps 1. Color based segmentation The color segmentation generates a binary mask with the same size of the original image. Using binary image with white regions showing skin regions and the black area showing non-skin regions, each skin region must be labeled.

Page 2

International Journal of Communications and Engineering Volume 01 No.1, Issue: 03 March2012

2. FIXED TEMPLATE MATCHING A template of size of frame is created. Here we assume that face will be present in the centre of frame. So centre of template will be a white square of size of face, surrounded by Black Square. This template is AND operated with color based segmentation output. 3. Morphological operation After fixed template matching the left over noise in the background can be smoothened using morphological processing 4. Find Euler Number 5. Determine number of Holes The segmented skin region is now evaluated assuming that the face would consists of at least one hole as it consists of eyes, nose and a mouth. If the hole is absent then it can be treated pseudo skin pixel or area. H=C-E Where E is Euler number determined by bweuler function in MATLAB, C is number of connected component. It value is equal to 1, as only one segmented region is considered at a time. H is number of holes. 6. When number of holes>0, determine whether width/height of segment lies in the interval 0<ratio<4 7. If ratio is satisfied, the portion is detected as face, else not a face 3 FACE RECOGNITION Face recognition is the second phase. It involves comparing the input face with faces in the database. PCA algorithm is used for recognizing face (6) (7). Photos

in the database is called training images and the photo taken during authentication phase is called as test image. In the PCA algorithm that we use to implement authentication system, an offline tuning is done after enrolment phase, to determine the threshold. In PCA algorithm a weight vector is computed for all images in the database as well as for the new test image. Test image matches with that image in the database if minimum Euclidian distance between corresponding weight vectors is less than particular threshold. To find out this threshold offline tuning is done. Here each training image is given as test image and minimum Euclidian distance is computed. Thus we get a vector of Euclidian distances with number of elements equal to number of photos in the database. Algorithm for determining threshold using this vector is explained in following section One of serious issue effecting face recognition in 2D is unequalized illumination. To avoid this problem, before applying PCA algorithm, all the extracted faces in database will undergo DCT normalization. DCT is employed in the logarithm domain. Since illumination variation lies mainly in the low frequency band, an appropriate number of DCT coefficients are truncated to reduce the variations under different lighting conditions. Block diagram of face recognition is shown in Fig.

Fig 3.1 Block diagram of face recognition

Page 3

International Journal of Communications and Engineering Volume 01 No.1, Issue: 03 March2012

Before going into detailed explanation of algorithm, we will see how this algorithm as a whole works. The algorithm can be divided into two stage 1) Enrolment Phase 2) Authentication Phase. In enrolment phase, photos of all drivers who owner wishes to authenticate will be taken and face portion alone will be extracted from those photos to create database. In enrolment phase, we will compute weight vectors for all images in database using PCA algorithm. This weight vector specifies, to what degree the specific feature is present in the original image. Once this is computed, authentication phase starts. In this phase, we will capture the image of person entering vehicle and do image processing on it. Image processing means applying PCA algorithm on it. As in enrolment phase, here also we will compute the weight vector of this test image. Calculate the Euclidian distance between weight vector of test image and each of weight vectors of database images. If new driver is an authenticated driver, then minimum Euclidian distance will be below certain threshold. If minimum Euclidian distance is greater than this threshold, the new driver is not an authenticated one. If it is less than this threshold, the new driver is an authenticated one. 3.1 Algorithm of Face Recognition (6) (7) 1. Get training set of images 2. Find mean of images 3. Find difference between mean image and each of training images. 4. Find covariance matrix of the matrix obtained from step 2 5. Find Eigen values and Eigen vectors of this covariance matrix

6. Find Eigen faces with larger Eigen values 7. Find out weight vector using this Eigen faces 8. Give new test image 9. Find out weight vector for test image 10. Find Euclidian distance between weight vectors of test image and training images 11. If this distance is less than threshold then test image is considered to be in database and hence authenticated, otherwise not authenticated. 4 THRESHOLD DETERMINATION ALGORITHM 1. Find minimum Euclidian distance, by giving each training image as test image. Let the vector of minimum Euclidian distances be E. 2. Find mean of E. Let it be M1 3. A=E-M1 4. Square each terms of A. Denote it as B. 5. M2=mean (B) 6. Standard deviation=sqrt (M2) 7. Standard deviation*2.2=C 8. Threshold=C+M1 5 GSM/GPRS MODEMS GPRS can be used for services such as Wireless Application Protocol (WAP) access, Short Message Service (SMS), Multimedia Messaging Service (MMS), and for Internet communication services such as email and World Wide Web access. GPRS data transfer is typically charged per megabyte of traffic transferred, while data communication via traditional circuit switching is billed per

Page 4

International Journal of Communications and Engineering Volume 01 No.1, Issue: 03 March2012

minute of connection time, independent of whether the user actually is using the capacity or is in an idle state. GPRS is a best-effort packet switched service, as opposed to circuit switching, where a certain Quality of Service (QoS) is guaranteed during the connection for nonmobile users. 6 GPS RECEIVER A GPS receiver receives radio signals from 24 satellites orbiting the Earth. If the receiver can receive a signal from at least 3 satellites simultaneously, it can calculate its own location, usually accurate to within a few feet. At this point, this location information must somehow be transmitted to a base station to be displayed on a computerized map. The GPS module can boot up in one of three modes: Hot Start, Warm Start or Cold Start. The Time-To-First-Fix (TTFF) depends on the startup mode, with cold starts giving the longest TTFF. The almanac contains satellites orbit information and allows the GPS receiver to predict which satellites are overhead, shortening acquisition time. The GPS receiver must have a valid almanac to be capable of booting up in warm or hot start modes. The receiver must have a continuous fix for approximately 15 minutes to receive a complete almanac from the satellites. Once downloaded it is stored in nonvolatile memory. Execution of a cold start will automatically result in a new almanac download. Ephemeris data contains precision corrections to the almanac data and is required for accurate positioning. 7 ENTIRE SYSTEM ALGORITHM 1. Get training set of images 2. Detect face of all training images 3. Do DCT normalization for uniform illumination

4. Input these images as training set into face recognition stage 5. Compute weight vector for input training set using PCA algorithm 6. Give new test image 7. Determine whether new image is authenticated or not using PCA algorithm 8. If driver is authenticated, switch on relay which is connected between ignition coil and battery and thus give access to vehicle 9. If driver is not authenticated, the captured image is send to the owners mobile. 10. If the owner doesnt knew the driver means the owner can stop the car by sending the secret code to the GSM Modem. 8 CONCLUSION Even though there are lots of car alerts or security product but the number of car theft still mounting hurriedly. The proposed vehicle monitoring, alert, and tracking model presents the good easy-tobuild system possible. In this paper, it is clear fact that the system is more portable and less expensive system. It is an efficient model in the sense that it uses mobility capabilities like MMS for alerting the owner. This vehicle security system gives the perfect solution for many related problems such as, organizing the stolen cars cases, notification speed, data and information storage and accuracy. The implementation and testing results for the system were good, the process of sending MMS messages was successful, the process of receiving the MMS by the end users and the database was successful also, and it took an average time about 40 seconds to get alerted the car owner. This short time is mostly enough

Page 5

International Journal of Communications and Engineering Volume 01 No.1, Issue: 03 March2012

for the owner to reach his car before the intruder taken the car, especially if the owner is nearby his car. The database will receive the intruder photo within 4 minutes, which is very helpful for the security and police to take action against the intruder. This will leads to reduce car theft cases and increase the possibility to find stolen cars and intruders. 9 REFERENCES
1. A. Pazhampilly Sreedevi, B. Sarath S Nai Image Processing Based Real Time Vehicle Theft Detection And Prevention System,2011 2. Shihab A. Hameed, Shaima Abdulla, Mohd Ershad, Fauzan Zahudi, Aisha Hassan New Automobile Monitoring and Tracking Model 2011 4th International Conference on Mechatronics (ICOM), 17-19 May 2011, Kuala Lumpur, Malaysia 3. Jian Xiao and Haidong Feng A Low-cost Extendable Framework for Embedded Smart Car Security System, IEEE International Conference on Networking, Sensing and Control, Okayama, Japan, March 26-29, 2009 4. Joseph A. O'Sullivan, Robert Pless, Advances in Security Technologies: Imaging, Anomaly Detection, and Target and Biometric Recognition, Microwave Symposium IEEE/MTT-S International Volume, Page(s):761 764, 2007. 5. S. Ajaz, M. Asim, M. Ozair, M. Ahmed, M. Siddiqui, Z. Mushtaq, Autonomous Vehicle Monitoring & Tracking System, SCONEST 2005, pp. 1 4, 2005.

6.M.Turk. and A.Pentland., Face recognition using eigenfaces, Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, Maui, Hawaii, pp. 586-591, 3-6 June 1991. 7.M. Turk, A. Pentland, Eigenfaces for Recognition, Journal of Cognitive Neurosicence, Vol. 3, No. 1, pp. 71-86, 1991

Page 6

Das könnte Ihnen auch gefallen