Project_Report_Group - Abijith

Please enable JavaScript to view the full PDF

SOCIAL NETWORK OF SAFE VEHICLES Project Report Submitted in partial fulfillment for the award of the degree of BACHELOR OF TECHNOLOGY In COMPUTER SCIENCE & ENGINEERING (VIIth Semester) 2015 - 2019 Batch APJ Abdul Kalam Technological University By ABIJITH C G (SHR15CS002) AJITH K ASHOK (SHR15CS004) AKSHAY B NAIR (SHR15CS007) NOVEMBER 2018 DEPARTMENT OF COMPUTER SCIENCE & ENGINEERING SAHRDAYA COLLEGE OF ENGINEERING AND TECHNOLOGY, KODAKARA, THRISSUR BONAFIDE CERTIFICATE This is to certify that the project report titled ”SOCIAL NETWORK OF SAFE VE- HICLES” is bonafide work of ABIJITH C G (SHR15CS002), AJITH K ASOK (SHR15CS004) & AKSHAY B NAIR(SHR15CS007) during their VIIth semester B.Tech in partial fulfillment of the requirements of the APJ Abdul Kalam Tech- nological University, under our supervision. COORDINATOR PROJECT GUIDE & HEAD OF THE DEPARTMENT Ms. Priya K V Mr. Krishnadas J Assistant Professor Assistant Professor Kodakara 26-11-2018 ACKNOWLEDGMENT We would like to express our immense gratitude and profound thanks to all those who helped us to make this project a great success. We express our gratitude to the almighty God for all the blessings endowed on us. We take this opportunity to express our sincere thanks to our Executive Director Rev. Fr. George Pareman and Principal Dr. Nixon Kuruvila for providing us with such a great opportunity. We also convey our gratitude to our Head of the Department Mr. Krishnadas J for having given us a constant inspiration and suggestion. We extend our deep sense of gratitude to our project coordinator Ms. Priya K V, Assistant Professor of Com- puter Science & Engineering Department for providing and enlightening guidance through the project. We can hardly find words to express our deep appreciation for the help and warm encouragement that We have received from our project guide Mr. Krishnadas J, Assistant Professor of Computer Science & Engineering Depart- ment for his whole-hearted support. It was their encouragement that helped us to complete the project. We can hardly find words to express our deep appreciation of the help and warm encouragement that We received from our parents. We are extremely thankful and indebted to our friends who supported us in all aspects of the project work. i INSTITUTIONAL VISION To train the youth to be the leaders of tomorrow with apt skills, deep rooted sense of social responsibility, strong ethical values and with a global outlook to face the challenges of changing world. INSTITUTIONAL MISSION With a high calibre faculty and an excellent infrastructure, we promote academic excellence, absolute discipline and sound practical exposure. QUALITY POLICY We at Sahrdaya are committed to provide Quality Technical Education through con- tinual improvement and by inculcating Moral & Ethical values to mould into Vibrant Engineers with high Professional Standards. We impart the best education through the support of competent & dedicated facul- ties, excellent infrastructure and collaboration with industries to create ambience of excellence. ii DEPARTMENTAL VISION To evolve as a national level Center of Excellence in academics and to research with the aim of imparting contemporary knowledge in the field of Computer Science and Engineering. DEPARTMENTAL MISSION 1. Have state of art infrastructure and resources for teaching and research. 2. Impart relevant technical knowledge,skills and attributes along with values and ethics. 3. Enhance research quality and creativity through innovative teaching learn- ing methodologies. 4. Mold Computer Science Engineering Professionals in synchronization with the dynamic industry requirements, worldwide. 5. Inculate essential leadership qualities coupled with commitment to the so- ciety. PROGRAMME EDUCATIONAL OBJECTIVES (PEOS) PEO1 Take up challenging careers in suitable corporate, business or ed- ucational sectors across the world, in multi-cultural work environ- ment. PEO2 Continuously strive for higher achievements in life keeping moral and ethical values such as honesty, loyalty, good relationship and best performance, aloft. PEO3 Be knowledgeable and responsible citizens with good team-work skills, competent leadership qualities and holistic values. iii PROGRAMME SPECIFIC OBJECTIVES (PSOS) PSO1 To nurture students with technically inquisitive attitude so that any real- world problem could be tackled with a problem solving per- spective, finding a suitable mathematical model with strong funda- mental technological concepts to solve and apply to rapid growing arena of computer technology. PSO2 To develop professionals with excellent exposure to the latest tech- nologies to design high quality products unique in innovation, tech- nology, software, security, hardware and usefulness; making high impact on society, business and technology. PSO3 To enhance knowledge in practical implementation of technology with regard to parallelism, virtualization of networks, scientific analysis and modeling, visualization, natural language processing, digital synthesis of data and its manipulation, wireless and mobile communication, storage and retrieval of huge amount of data etc. PROGRAMME OUTCOMES (POS) PO1 Engineering knowledge: Apply the knowledge of mathematics, sci- ence, engineering fundamentals, and an engineering specialization to the solution of complex engineering problems. PO2 Problem analysis: Identify, formulate, review research literature, and analyze complex engineering problems reaching substantiated conclusions using first principles of mathematics, natural sciences, and engineering sciences. PO3 Design/development of solutions: Design solutions for complex engineering problems and design system components or processes that meet the specified needs with appropriate consideration for the public health and safety, and the cultural, societal, and environmen- tal considerations. iv PO4 Conduct investigations of complex problems: Use research-based knowledge and research methods including design of experiments, analysis and interpretation of data, and synthesis of the information to provide valid conclusions. PO5 Modern tool usage: Create, select, and apply appropriate tech- niques, resources, and modern engineering and IT tools including prediction and modeling to complex engineering activities with an understanding of the limitations. PO6 The engineer and society: Apply reasoning informed by the con- textual knowledge to assess societal, health, safety, legal and cul- tural issues and the consequent responsibilities relevant to the pro- fessional engineering practice. PO7 Environment and sustainability: Understand the impact of the pro- fessional engineering solutions in societal and environmental con- texts, and demonstrate the knowledge of, and need for sustainable development. PO8 Ethics: Apply ethical principles and commit to professional ethics and responsibilities and norms of the engineering practice. PO9 Individual and team work: Function effectively as an individual, and as a member or leader in diverse teams, and in multidisciplinary settings. PO10 Communication: Communicate effectively on complex engineering activities with the engineering community and with society at large, such as, being able to comprehend and write effective reports and design documentation, make effective presentations, and give and receive clear instructions. PO11 Project management and finance: Demonstrate knowledge and un- derstanding of the engineering and management principles and ap- ply these to one’s own work, as a member and leader in a team, to manage projects and in multidisciplinary environments. PO12 Life-long learning: Recognize the need for, and have the prepara- tion and ability to engage in independent and life-long learning in the broadest context of technological change. v COURSE OBJECTIVES To develop skills in doing literature survey, technical presentation and report prepa- ration. COURSE OUTCOMES The student will be able to CO1 Analyze a current topic of professional interest and present it before an audience CO2 Identify an engineering problem, analyze it and propose a work plan to solve it. CO3 Design a model with respect to recent technologies in the field of computer science. CO4 Describe, compare and evaluate different technologies CO-PO MAPPING PO1 PO2 PO3 PO4 PO5 PO6 PO7 PO8 PO9 PO10 PO11 PO12 CO1 3 3 CO2 3 2 2 CO3 3 2 CO4 2 #1 – Weak correlation 2- Moderate correlation 3- Substantial correlation CO-PSO MAPPING PSO1 PSO2 PSO3 CO1 2 CO2 3 CO3 3 CO4 2 3 #1 – Weak correlation 2- Moderate correlation 3- Substantial correlation vi PROJECT OUTCOMES PR1 Use acquired knowledge to examine solutions to Pothole Identifica- tion & Reduction. PR2 Distinguish and Integrate differing forms of knowledge. PR3 Apply the use of a Machine Learning to meet identified challenges. PR4 Apply principles of ethics and respect in interaction with others. MAPPING OF PRS WITH COS CO1 CO2 CO3 CO4 PR1 2 1 PR2 3 PR3 1 3 1 PR4 3 #1 – Weak correlation 2- Moderate correlation 3- Substantial correlation MAPPING OF SRS WITH POS PO1 PO2 PO3 PO4 PO5 PO6 PO7 PO8 PO9 PO10 PO11 PO12 PR1 1 1 2 PR2 2 1 2 PR3 1 1 3 2 3 PR4 2 3 1 1 1 #1 – Weak correlation 2- Moderate correlation 3- Substantial correlation MAPPING SR WITH PSOS PSO1 PSO2 PSO3 PR1 2 PR2 3 PR3 2 2 PR4 1 #1 – Weak correlation 2- Moderate correlation 3- Substantial correlation vii ABSTRACT A significant percentage of road accidents happening in India are mainly caused by poor road infrastructure. Potholes, cracks, uneven road-surfaces, unaware speed bumps and bad weather conditions all are dangerous traps which increases the possi- bility for an accident. Motorcycles are on the top of the list of vehicles that gets into deathly hazards due to these reasons. These problems are also have a deteriorating effect on the chassis of the vehicle. The driver, being unaware of the traps ahead can’t do much in order to escape from them. The current proposal is to warn the drivers about the poor road conditions ahead, by giving them enough time to take necessary action to prevent the hazard. The system also proposes to send the infor- mation regarding the bad conditions of the road to the respected authorities such as National Highways Authority, Public Works Department etc. Irrespective of their actions, the system will continue giving warnings,until the repairs are done. Once the necessary steps are taken, it will be automatically detected and corrected in the system. The information about the road conditions will be collected by analyzing the movements of the suspension systems of automobiles.Data from each vehicle will be collected and sent to the cloud where they will be analyzed, processed and optimized. Once the data is optimized, each vehicle will be safe to go, being alert about the possible hazards that awaits them in the future roads. viii CONTENTS ACKNOWLEDGMENT . . . . . . . . . . . . . . . . . . . . . . . . . . i INSTITUTIONAL VISION, MISSION AND QUALITY POLICY . . ii DEPARTMENTAL VISION, MISSION, PEOs ,PO AND PSOs . . . . iii ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viii LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi LIST OF ABREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . xii 1 INTRODUCTION 1 1.1 General Background . . . . . . . . . . . . . . . . . . . . . . . . . 1 1.2 Problem Definition . . . . . . . . . . . . . . . . . . . . . . . . . . 2 1.3 Motivation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 1.4 Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4 2 LITERATURE SURVEY 5 2.1 Vehicle Data Crowd sourcing . . . . . . . . . . . . . . . . . . . . . 5 2.2 Pothole Detection . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3 Road Bank and Incline Angles . . . . . . . . . . . . . . . . . . . . 7 3 SYSTEM ANALYSIS 8 3.1 Existing System . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8 3.2 Proposed System . . . . . . . . . . . . . . . . . . . . . . . . . . . 10 4 DESIGN AND DEVELOPMENT 12 4.1 Module Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 4.2 Module Description . . . . . . . . . . . . . . . . . . . . . . . . . . 13 4.3 Block Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 4.4 Flow Chart . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16 4.5 Data Flow Diagram . . . . . . . . . . . . . . . . . . . . . . . . . . 18 5 TECHNOLOGY DESCRIPTION 20 5.1 Hardware Specifications . . . . . . . . . . . . . . . . . . . . . . . 20 5.1.1 Accelerometer . . . . . . . . . . . . . . . . . . . . . . . . . 20 5.1.2 Gyroscope . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 5.1.3 Ultrasonic Sensor . . . . . . . . . . . . . . . . . . . . . . . 21 ix 5.1.4 Raspberry Pi . . . . . . . . . . . . . . . . . . . . . . . . . 22 5.2 Machine Learning Algorithms . . . . . . . . . . . . . . . . . . . . 22 5.2.1 C4.5 decision tree Classifier . . . . . . . . . . . . . . . . . 23 5.2.2 SVM . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 5.2.3 Naive Bayes . . . . . . . . . . . . . . . . . . . . . . . . . . 28 5.3 Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31 6 RESULTS & PRESENT STATUS 33 7 CONCLUSION & FUTURE WORK 37 REFERENCES 38 APPENDIX xiv I Base Paper 1 xiv II Base Paper 2 xxix III Base Paper 3 xxxviii LIST OF FIGURES 3.1 Pothole Patrol road monitoring architecture. . . . . . . . . . . . . . 9 4.1 Modular design diagram of proposed model . . . . . . . . . . . . . 13 4.2 Block diagram of proposed model . . . . . . . . . . . . . . . . . . 15 4.3 Flow chart 1 of proposed model . . . . . . . . . . . . . . . . . . . . 16 4.4 Flow chart 2 of proposed model . . . . . . . . . . . . . . . . . . . . 17 4.5 Level 0 DFD . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 4.6 Level 1 DFD. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 5.1 Example of a decision tree in flowchart form. . . . . . . . . . . . . 25 5.2 An example of a two-class problem with two separating hyperplanes, B1 and B2. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 6.1 Decision Tree . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 6.2 Test Cases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36 xi LIST OF ABBREVIATIONS ABBREVIATION DESCRIPTION WHO World Health Organisation SVM Static Vector Machine GPS Global Positioning System MEMS Micro-electromechanical System ANN Artificial Neural Networks LogR Logistic Regression CSV Comma Separated Values xii CHAPTER 1 INTRODUCTION Dangerous road surface conditions are major distractions for safe and comfortable transportation. Both drivers and road maintainers are interested in fixing them as soon as possible. However, these conditions have to be identified first. According to statistics provided by World Health Organization (WHO), road accidents have become one of the top 10 leading causes of death in the world. Specifically, road accidents claimed nearly 1.25 million lives per year (2015). Studies in show that most road accidents are caused by poor condition of roads. Bad roads are a big problem for vehicles and drivers, this is because the deterioration of roads leads to more expensive maintenance, not only for the road itself but also for vehicles. Accordingly, road surface condition monitoring systems are very important solutions to improve traffic safety, reduce accidents and protect vehicles from damage due to bad roads. Both road managers and drivers are interested in having sufficient information concerning road infrastructure quality (safe or dangerous road). 1.1 GENERAL BACKGROUND One approach to road damage detection is to use human reports to central authori- ties. While it has the highest accuracy, assuming that people are fair, it also has the most human interaction and is not comprehensive. Statistical analysis can be used to estimate damage probabilities of road segments based on their usage intensity. The simplest method might be to collect photos of road damage and hazards taken by the participants and to upload them to a central server. However, this requires strong participation and interaction from the users as well as manual image analysis. We believe that an automated approach for detecting potholes with little or no human interaction is more promising. This would ensure more comprehensive survey data with less errors caused by human factors than generated by mere enthusiasm of the participants. SOCIAL NETWORK OF SAFE VEHICLES An automated survey approach could be carried out by either customized embedded sensing devices or smart-phones. While the former has more sensing ca- pabilities and adaptation potential, the popularity of smart-phones makes the latter approach very appealing in terms of practical usability. To create a successful road surface monitoring system accepted by wide user community, it is important to make it attractive for the users, to provide added value without a significant process over- head. Therefore we envision our system as a service, which is added as a layer to existing navigation systems, such as Waze, which use real-time traffic information, collected by participatory sensing approach. Although contemporary smart-phones have high processing power and considerable memory, the detection system is rec- ommended to avoid resource-intensive detection methods and to preserve initial user interface responsiveness. In this ever fast moving world everyone looks to reach their destinations as early as possible. The major limitation to this dream is the improper, irregular roads. The drivers are unable to drive the vehicle with a regular speed due to the sudden existence of hazardous pothole and humps on the road. This also causes severe damage to the vehicle. Current road surface monitoring uses human effort to check the condition and quality of the road which makes the process more time consuming and less efficient. An automatic system which is capable of detecting road anomalies without any human effort may completely transform the current road transportation system efficiency. Identification of pavement distress such as potholes and humps not only helps drivers to avoid accidents or vehicle damages, it also helps authorities to keep track of the road conditions for its better maintenance. 1.2 PROBLEM DEFINITION The importance of the road infrastructure for the society could be compared with importance of blood vessels for humans. To ensure road surface quality it should be monitored continuously and repaired as necessary. The optimal distribution of resources for road repairs is possible providing the availability of comprehensive and objective real time data about the state of the roads. Participatory sensing is a promising approach for such data collection. Monitoring the road condition has acquired a critical significance during recent years. There are different reasons behind broadening research on this field: to start with, it will guarantee safety and comfort to different road users; second, smooth streets will cause less damage to the car. Our motivation is to create a real- time Application System that automatically predicts the quality of the road based on tri-axial accelerometer, gyroscope and ultrasonic sensors, show the road location trace on a geographic map using GPS and save all recorded workout entries. Sahrdaya College of Engineering & Technology, Kodakara 2 SOCIAL NETWORK OF SAFE VEHICLES C4.5 Decision tree classifier is applied on training data to classify road segments and to build our model. Using this approach, we expect to visualize a road quality map of a selected region. Hence, we can provide constructive feedback to drivers and local authorities. Besides, Road Manager can benefit from this system to evaluate the state of their road network and make a checkup on road construction projects, whether they meet or not the required quality. 1.3 MOTIVATION Nowadays various piece of data can be used to acquire necessary information. Par- ticularly we refer to processing of road data collected from 3-axis acceleration sensor and GPS (Global Positioning System). This technical data provides important clues which serve to understand different patterns occurred during the driving. As a goal we set to study distinct behaviors of accelerometer records with respect to road con- ditions and some particular driving events. Given research facilitates road quality identification. As a matter of fact, the problem of quality of roads still remains open. The maintenance of roadways calls for regular monitoring of road degrees, which can be classified as smooth and non-smooth. Smooth roads are divided into several types such as perfectly flat surface, smooth but rough and smooth road with some anomalies. These anomalies are damaged parts like pits and potholes. Some physical damages can be caused because of the weather conditions. Non-smooth is also defective type of the road which comprises uneven surface, poorly paved and non-paved roads. The latter one we also name as off-roads. Most of these road issues have their impacts on state of the vehicles. Especially they deteriorate the chassis (vehicle frame) of an automobile. They can also result in defects of the car. To prevent something like this, drivers need to be notified about bad road or abnormality so they can bypass them. In order to reconstruct damaged part of the road in a short period of time maintenance services must be informed of them instantly. Such awareness can be achieved through the system which will automatically analyze road condition. To offer the solution one need to understand specificity of different patterns on the road. Precisely we only need to explore 3-axis accelerometer values and GPS coordinates provided by sensors built into a device which is attached inside the vehicle. So the idea is that acceleration changes can give distinguishable data in various situations. Using GPS coordinates we store traveling position and calculate average speed at a certain time. Regardless of the practical information we still face some challenging patterns and noisy data. Sahrdaya College of Engineering & Technology, Kodakara 3 SOCIAL NETWORK OF SAFE VEHICLES 1.4 OBJECTIVES The objectives of this work are: • The primary objective is to ensure safety of passengers, improve the road con- ditions and adapt to the changing conditions quickly as possible by informing the authorities. • Reduce the number of accidents that happen due to poor road condition. • Make drivers alert of the road infrastructure and weather condition, so as to take decision on what route to proceed with. • Prevent deterioration of vehicle’s chassis, caused by deep potholes and uneven road surfaces. • Alert authorities about the critical maintenance’s. • Make night drives hassle-free. Sahrdaya College of Engineering & Technology, Kodakara 4 CHAPTER 2 LITERATURE SURVEY In recent years, road condition monitoring has become a popular research area. There has been some considerable works done in this field. 2.1 VEHICLE DATA CROWD SOURCING Crowd sourcing information, and the architectural design of such systems [7], has been investigated in prior works for multiple vehicular purposes. Examples include locating available road-side parking spots [8], estimating travel time and mapping GPS traces to road segments [9], identifying anomalous traffic states from GPS trace data [10], generating digital maps from vehicle GPS traces and distinguishing be- tween one-way and two-way roads , and using GPS data to identify traffic lights and stop signs . In some of these scenarios, the vehicle sensors are able to directly mea- sure the feature-of-interest (e.g., travel time, direction of travel). They therefore do not consider the different possible mappings from sensor data to an event that may occur, such as in our work where different accelerometer patterns from different ve- hicles all correspond to the same pothole. There is an increased challenge in crowd sourced detection when the mapping from the observed signals to the detected fea- ture is not identical for all vehicles. Methods of aggregating heterogeneous or conflicting information have been in- vestigated. The work in uses a Time-Decay Sequential Hypothesis Testing algorithm to aggregate and disseminate road information to and from multiple vehicles. How- ever, at low sampling rates this relies on aggregating decisions from weak classifiers instead of using the collected raw data. Each weak classifier is unreliable due to its limited access to data, which results in inaccurate detection’s when aggregating the decisions. SOCIAL NETWORK OF SAFE VEHICLES 2.2 POTHOLE DETECTION Pothole detection systems have been the focus of previous works, however the sensor device has generally been a smart phone instead of embedded vehicle sen- sors. We use embedded sensors since they are standardized across vehicles and are integrated within the controls and communications systems. This distinction is fun- damental to this work since smart phones generally have a much higher sensor op- erating frequency (300+ Hz) than embedded vehicle sensors. These higher frequen- cies allow for a detection system to measure the full dynamic motion of the vehicle caused by road features. This differs from the low frequency acceleration signals in our work where some of the signal properties that distinguish pothole regions in a single vehicle are lost since so little data is available. Signal under sampling has been shown to be problematic for other vehicular applications such as speed estimation . It is therefore difficult to perform any comparison between these methods since, as will be shown, the sampling frequency is a principal dictator in the types of available features. Regarding these existing pothole detection methods, the Pothole Patrol system uses speed, high-pass, and vertical and lateral acceleration filters to identify potholes from test signals from Boston taxis. Road bumps are detected in by examining the peak vertical acceleration and the duration for which the acceleration dips below a heuristically defined threshold. Gaussian Mixture Models are used in on aggregated data to determine potholes from 100 taxis in Shenzhen, China by examining z-scores of the listed Pothole Patrol features . A linear model for speed using 38 Hz sensors was constructed to try to eliminate the speed dependence in, however the vertical acceleration, which is used to create most of the features, can deviate significantly from the linear model. Since these works use non-embedded high-frequency sensors, they rely on only one vehicle for detection. They therefore ignore problems resulting from GPS error, which increases classification difficulty. This is particularly problematic when ag- gregating data from multiple vehicles with different dynamics as the exact ordering of the measurements cannot be determined. These, and previous works which use crowd sourcing detection techniques for pothole detection , do not consider multi- lane scenarios where GPS position error obfuscates the pothole data with normal road data from adjacent lanes. We also address the issue of finely localizing the pot- hole longitudinally on the road, which has not been addressed in previous works. There are other systems which attempt to help drivers by pre-detecting potholes in real-time by using images from vehicle-mounted cameras instead of accelerome- ters. Sahrdaya College of Engineering & Technology, Kodakara 6 SOCIAL NETWORK OF SAFE VEHICLES However, even with the advantage of being able to picture the pothole before the vehicle enters it, the required angle of the camera and subsequent pro- cessing time does not necessarily provide the driver the opportunity to avoid the pothole. Also, unlike accelerometer-based detection methods, image-based methods require specific lighting conditions to function properly and may be unusable at night or in poor weather conditions. With regard to the image based methods, the work in seeks pothole regions in images via segmentation methods employing shape-based thresholding, and examining geometric and texture based properties of suspect pot- hole regions. Similarly, large simulated potholes are found in images in by looking for large circular objects with a predetermined brightness difference. 2.3 ROAD BANK AND INCLINE ANGLES Vehicles have acceleration components unrelated to potholes, such as from grav- ity or centripetal acceleration, which can be determined in the vehicle reference frame from knowledge of the bank and incline angles of the roads. By estimat- ing these angles, the acceleration components unrelated to potholes can be filtered from the detection system, thus reducing the bandwidth required for transmitting data from the vehicles to the Cloud. This is a more systematic filtering method than the Mahalanobis distance thresholding on acceleration components in which could still allow large acceleration components through the filter that may be expected from normal driving conditions. Our road angle calculations are adapted from , and each angle calculation uses only a single accelerometer reading and corresponding GPS measurement. There have been other approaches to determine this road angle information, however they generally rely on significantly more variables and vehicle- specific parameters. High frequency sensors are used to estimate vehicle roll and pitch angles using inertial sensor data in . In , known four-wheel speeds, steering angles, acceleration, and gyroscopic data are used to construct a non-linear road-tire friction model to estimate road inclination and bank angles. Vehicle mass, gear and engine torque, wheel radius, vehicle frontal area, and air constant parameters are used with inertial sensor data to design a Kalman filter to estimate the road incline. Some works use multiple GPS devices to identify road angles . The carrier phase difference between two roof mounted GPS antennae as well as the ratio of the cal- culated horizontal and vertical velocities are used to determine the road grade and vehicle mass. Using other sensors, a barometer in addition to acceleration and gyro- scope data is used to estimate the road inclination. Our system does not rely on such extensive measurements; only GPS, ultrasonic sensors and accelerometer measure- ments are required. Sahrdaya College of Engineering & Technology, Kodakara 7 CHAPTER 3 SYSTEM ANALYSIS The system introduces a road condition monitoring framework which is based on sensors (accelerometer, gyroscope and GPS) built in smart phones to give us the quality of different road sections using machine learning techniques. The contribu- tions are manifold and can be summarized as follows: • As a first contribution, we design a machine-learning algorithm (C4.5 Decision tree) to classify road segment as compared to previous works that use simple thresholds, SVM and fuzzy logic. Our tests show that our system is able to detect and classify events related to road conditions with an accuracy of 98,6%. • The system, is an inexpensive simple yet efficient solution that is able to mon- itor road quality. It is realized on Android smart phones and is highly portable and easy to maintain. The application provide constructive feedback to drivers and local authorities by plotting the evaluated road location on a Map and saving all recorded workout entries. • Creating an Android application that allows real-time and automatic collection and analysis of accelerometer and gyroscope data in order to get reliable road surface labels in contrast to previous works that mostly use offline methods (videos, images for data labeling). • While most of previous works employ uni modal accelerometer data, using gyroscope sensor in conjunction with accelerometer sensor will derive more accurate road quality prediction. 3.1 EXISTING SYSTEM The Pothole Patrol [4] is a sensing application that reports the road surface condi- tions. It require the integration of particular hardware equipment; for each vehicle an embedded computer running Linux is used for data processing, a Wi-Fi card for transmitting gathered data, an external GPS for localization, and a 3-axis ac- celerometer to monitor road surface. It uses a specific hardware/software platform SOCIAL NETWORK OF SAFE VEHICLES - Linux powered Soekris 4801 embedded computers with external accelerometers (sampling rate 380Hz) and an external GPS. Their pothole detection algorithm is based on simple machine-learning approach using X and Z axis acceleration and the vehicle velocity data as input. The algorithm consists of five consecutive filters: speed, high − pass, z − peak, xz − ratio and speed vs z − ratio. Each filter is used as a rejecter of one or more event types not related to potholes such as door slams or railway crossings. Additional training process is executed for optimal tuning of the last three filters. This system had been deployed on 7 taxis driving in Boston. The system is able to detect road anomalies and distinguish them from other unnecessary driving events. Pothole Patrol system with false positive reduction approach was able to identify potholes in need of repair with 90% precision. The system can also determine the location of found potholes on the map. It is done by analyzing GPS coordinates of the detected road anomalies. Fig. 2.1 shows the P2 architecture[4]. Fig. 3.1: Pothole Patrol road monitoring architecture. Nericell [3] is a system developed by Microsoft to monitor roads and traf- fic conditions. It requires a very complicated hardware and software setup. It uses several external sensors such as a microphone, GPS, Sparkfun WiTilt accelerome- ter. It uses using Windows Mobile OS powered smart-phones as hardware/software platform with an array of external sensors such as accelerometers (sampling rate 310Hz), microphones and GPS. Their algorithms for pothole detection z-sus (for speeds <25km/h) and z-peak (for speeds ≥ 25km/h) are based on simple threshold- based heuristics. Additional algorithm virtual reorientation is used to compensate arbitrary orientation of the smart-phone during driving in the vehicle. The detection is not very accurate (False positive rate less than 10% and false negative rate between 20% and 30%), the system may confuse between smooth, uneven and rough roads. Mednis et al. [5] proposed a real time system for detecting potholes. The system employs Android OS based smart phones having accelerometer sensor and simple algorithms to detect events from acceleration data. They provide detailed compar- ison of four methods Z-THRESH, Z-DIFF, STDDEV(Z) and G-ZERO on gathered Sahrdaya College of Engineering & Technology, Kodakara 9 SOCIAL NETWORK OF SAFE VEHICLES accelerometer data. In pothole detection algorithm Z-THRESH, events are repre- sented by measurements with values exceeding specified threshold level. In Z-DIFF events are represented by consecutive measurements with difference value above specific threshold level. STDDEV uses measurements with standard deviation value above specific threshold level. In G-ZERO events are represented by tuple of mea- surements with all three axis values below specific threshold level. Experimental results show a true positive rate equal to 90%. The drawbacks of this work is that the system uses only accelerometer sensor and data are collected through specialized hardware. There are some works that introduce own methods of measuring road quality. In [1] Kalra et al uses smart phone sensors to analyze driving events and road anomalies. They used accelerometer of smart phone placed onto dashboard to collect road and driving data. These data were observed and analyzed by determining thresholds for different patterns. Eriksson et al [2] proposed a system called Pothole Patrol which uses tree-axis accelerometer and GPS sensors for road data gathering. This system had been deployed on 7 taxis driving in Boston. The system is able to detect road anomalies and distinguish them from other unnecessary driving events. Pothole Patrol system with false positive reduction approach was able to identify potholes in need of repair with 90% precision. This system can also determine the location of found potholes on the map. It is done by analyzing GPS coordinates of the detected road anomalies. They also propose several filtering steps which dis- card events such as turning, braking, sensor orientation change, door slams and slow speed motion of a vehicle. Support Vector Machine was used to train classifiers on road quality data for event detection. Their approach is able to identify driving events like brakes and bumps. For bump detection it demonstrates a zero false pos- itive rate, and a 10% false negative rate. In previous paper [6], a design of system which uses a device with gyroscope and GPS sensors for road quality analysis was presented 3.2 PROPOSED SYSTEM As smart vehicles have become more ubiquitous, the capability now exists to detect environmental road features (e.g. potholes etc.) from their embedded sensor data. By aggregating data from multiple vehicles, crowd sourcing can be leveraged to detect environmental information with improved accuracy. We can thus focus on using such data to detect and localize potholes and analyze real time condition of roads. We focus on using real time continuous data from those sensor to classify the road condition and to provide real time warning to other vehicles passing through the same location. Thus we make use of different machine learning classifier to classify Sahrdaya College of Engineering & Technology, Kodakara 10 SOCIAL NETWORK OF SAFE VEHICLES the road condition data from the smart phone sensor into corresponding labels. The usage of data from the tri-axial accelerometer with machine learning techniques, we expect to visualize road condition of a certain region, hence providing useful feedback to the drivers and also to the local authorities to take measures to fix the road. We evaluate our system on real world data on different classifier to find out accuracy and performance of each cases 1. Uses the vehicle’s suspension system as the source of measurement. 2. Displacement of shock-absorbers is monitored continuously. 3. Linear Position Sensor gives data about the position of the shock-absorber. 4. When an unusual displacement is observed, it will be reported. 5. GPS data of the location-showing-unusual-reading, will also be collected. 6. During each specific time-interval, all the collected data will be sent to the Cloud server. 7. Data from each car will be sent to the cloud in this way. 8. Using statistical methods and optimization techniques, the bulk data will be analyzed in order to reach an inference. 9. These optimized inferences will be used to generate a road- map detailing the conditions of the roads. 10. Displacement of shock-absorbers is monitored continuously. 11. It will contain information such as GPS locations of potholes and speed-bumps, road infrastructure, optimal speed, shortest path to destination, weather condi- tion etc. 12. The driver will be alert of the possible hazards ahead, having enough time to take necessary actions. Sahrdaya College of Engineering & Technology, Kodakara 11 CHAPTER 4 DESIGN AND DEVELOPMENT The proposed system uses the vehicle’s suspension system as the source of mea- surement. There may be variations in the suspension system of vehicles. Thus the displacement of shock-absorbers is monitored continuously. This is done in order to unify the data values that are obtained from each vehicle. Even though the dis- placement of each vehicles shock absorber may vary, the relative change in the dis- placement will be very low. Linear Position Sensor gives data about the position of the shock-absorber. Hence when an unusual displacement is observed, it will be reported. GPS data of the location-showing-unusual-reading, will also be collected. During each specific time-interval, all the collected data will be sent to the Cloud server. Data from each vehicle will be sent to the cloud in this way. Using analyti- cal, statistical methods and optimization techniques, the bulk data will be processed and analyzed in order to reach an inference. These optimized inferences will be used to generate a road map detailing the conditions of the roads. Displacement of shock-absorbers is monitored continuously. By setting a threshold value we can thus distinguish between a pothole or hump. This threshold can be found out by taking the standard deviation value from the initial sets of data collected from the vehicles. It will contain information such as GPS locations of potholes and speed-bumps, road infrastructure, optimal speed, shortest path to destination, weather condition etc. The driver will be alert of the possible hazards ahead, having enough time to take neces- sary actions. 4.1 MODULE DESIGN Our goal is to derive a road quality recognition system that detects, analyzes, iden- tifies and predicts the state of road segments using ultrasonic sensors. Our system depends on pre-deployed infrastructures and additional hardware. In our system, road conditions could be detected and identified by ultrasonic sensors according to readings from accelerometer and gyroscope sensors. SOCIAL NETWORK OF SAFE VEHICLES Fig. 4.1: Modular design diagram of proposed model 4.2 MODULE DESCRIPTION The system is divided into two main modules, mainly the portion with the vehicle and that one in the main cloud server. The vehicular module is again subdivided into the following sections. • GPS: It is used to track the vehicles exact location i.e. when and where the event happened or occured. The Global Positioning System (GPS) is a net- work of about 30 satellites orbiting the Earth at an altitude of 20,000 km. Wherever you are on the planet, at least four GPS satellites are ‘visible’ at any time. Each one transmits information about its position and the current time at regular intervals. These signals, travelling at the speed of light, are intercepted by your GPS receiver, which calculates how far away each satellite is based on how long it took for the messages to arrive. Once it has information on how far away at least three satellites are, your GPS receiver can pinpoint your location using a process called trilateration. GPS satellites have atomic clocks on board to keep accurate time. General and Special Relativity however pre- dict that differences will appear between these clocks and an identical clock on Earth. General Relativity predicts that time will appear to run slower under stronger gravitational pull – the clocks on board the satellites will therefore seem to run faster than a clock on Earth. Furthermore, Special Relativity pre- dicts that because the satellites’ clocks are moving relative to a clock on Earth, they will appear to run slower Sahrdaya College of Engineering & Technology, Kodakara 13 SOCIAL NETWORK OF SAFE VEHICLES • Accelerometer: An accelerometer is an electromechanical device used to mea- sure acceleration forces. Such forces may be static, like the continuous force of gravity or, as is the case with many mobile devices, dynamic to sense move- ment or vibrations. Acceleration is the measurement of the change in velocity, or speed divided by time. Most accelerometers are Micro-Electro-Mechanical Sensors (MEMS). The basic principle of operation behind the MEMS ac- celerometer is the displacement of a small proof mass etched into the silicon surface of the integrated circuit and suspended by small beams. • Gyroscope: A gyroscope is a device that uses Earth’s gravity to help deter- mine orientation. Its design consists of a freely-rotating disk called a rotor, mounted onto a spinning axis in the center of a larger and more stable wheel. The accelerometer measures the acceleration + g (gravitational acceleration). Its functioning is based on elasticity of very tiny semiconductor beam inside the IC. Deflection of this beam is measured using piezoelectricity ( in most accelerometers). Gyro measures speed of angular rotation about an axis. • Linear Position Sensor: Linear position sensors convert the position into a proportional analog or digital signal. The output from the microprocessor is converted to a signal required by the user interface such as voltage, current, PWM or digital output. • Console: It simply refers to display where the driver can see the integrated map details of the route and various potholes and other hazards that are present along the route for the journey. It’ll either be put inside the vehicle as a sep- arate component or it can be integrated as an real-time application running in Mobile phones. Sahrdaya College of Engineering & Technology, Kodakara 14 SOCIAL NETWORK OF SAFE VEHICLES 4.3 BLOCK DIAGRAM Fig. 4.2: Block diagram of proposed model Sahrdaya College of Engineering & Technology, Kodakara 15 SOCIAL NETWORK OF SAFE VEHICLES 4.4 FLOW CHART Fig. 4.3: Flow chart 1 of proposed model Sahrdaya College of Engineering & Technology, Kodakara 16 SOCIAL NETWORK OF SAFE VEHICLES Fig. 4.4: Flow chart 2 of proposed model Sahrdaya College of Engineering & Technology, Kodakara 17 SOCIAL NETWORK OF SAFE VEHICLES 4.5 DATA FLOW DIAGRAM Fig. 4.5: Level 0 DFD Sahrdaya College of Engineering & Technology, Kodakara 18 SOCIAL NETWORK OF SAFE VEHICLES Fig. 4.6: Level 1 DFD. Sahrdaya College of Engineering & Technology, Kodakara 19 CHAPTER 5 TECHNOLOGY DESCRIPTION 5.1 HARDWARE SPECIFICATIONS 5.1.1 Accelerometer An accelerator looks like a simple circuit for some larger electronic device. De- spite its appearance, the accelerometer consists of many different parts and works in many ways, two of which are the piezoelectric effect and the capacitance sensor. The piezoelectric effect is the most common form of accelerometer and uses mi- croscopic crystal structures that become stressed due to accelerative forces. These crystals create a voltage from the stress, and the accelerometer interprets the volt- age to determine velocity and orientation. The capacitance accelerometer senses changes in capacitance between micro structures located next to the device. If an accelerative force moves one of these structures, the capacitance will change and the accelerometer will translate that capacitance to voltage for interpretation. Accelerometers are made up of many different components, and can be purchased as a separate device. Analog and digital displays are available, though for most technology devices, these components are integrated into the main technology and accessed using the governing software or operating system. Typical accelerome- ters are made up of multiple axes, two to determine most two-dimensional movement with the option of a third for 3D positioning. Most smart phones typically make use of three-axis models, whereas cars simply use only a two-axis to determine the mo- ment of impact. The sensitivity of these devices is quite high as they’re intended to measure even very minute shifts in acceleration. The more sensitive the accelerom- eter, the more easily it can measure acceleration. 20 SOCIAL NETWORK OF SAFE VEHICLES There are different types of accelerometers used in phones: 1. Piezoelectric Accelerometer: This device relies on the natural structures of piezoelectric crystals, which react to forces exerted on the phone by generating an electrical charge, which subsequently creates a voltage. 2. Micro-electromechanical System (MEMS): These are tiny mechanical struc- tures that change when forces are applied to them, subsequently changing an electrical property. 3. Capacitive Accelerometer: This device is a kind of MEMS. A net force applied to the mechanical system results in a change in the system’s capacitance. 5.1.2 Gyroscope A gyroscope is a wheel or disk mounted to spin rapidly about an axis and also free to rotate about one or both of two axes perpendicular to each other and to the axis of spin so that a rotation of one of the two mutually perpendicular axes results from application of torque to the other when the wheel is spinning and so that the entire apparatus offers considerable opposition depending on the angular momentum to any torque that would change the direction of the axis of spin. The sensors data of road surface quality were collected using accelerometer and gyroscope sensors built in the mobile phone, along the vehicle path. Several data collection drives were performed with a varied speed, the road condition label is pre-set before the collection starts. A typical mobile device has an accelerometer that can detect accelera- tion on two or three axes, allowing it to sense motion and orientation. A three- dimensional accelerometer can calculate pitch and roll and can be used in flight or driving simulation applications. Accelerometers consume a lot of energy, so they should be turned off when they are not being used to avoid draining a device’s bat- tery. 5.1.3 Ultrasonic Sensor Position Sensors detect the position of something which means that they are refer- enced either to or from some fixed point or position. These types of sensors provide a “positional” feedback. One method of determining a position, is to use either “dis- tance”, which could be the distance between two points such as the distance travelled or moved away from some fixed point, or by “rotation” (angular movement). For ex- ample, the rotation of a robots wheel to determine its distance travelled along the ground. Either way, Position Sensors can detect the movement of an object in a straight line using Linear Sensors or by its angular movement using Rotational Sen- sors. Sahrdaya College of Engineering & Technology, Kodakara 21 SOCIAL NETWORK OF SAFE VEHICLES 5.1.4 Raspberry Pi The Raspberry Pi is a series of small single-board computers developed in the United Kingdom by the Raspberry Pi Foundation to promote the teaching of basic computer science in schools and in developing countries. The original model became far more popular than anticipated, selling outside of its target market for uses such as robotics. Peripherals (including keyboards, mice and cases) are not included with the Rasp- berry Pi. Some accessories however have been included in several official and unof- ficial bundles. According to the Raspberry Pi Foundation, over 5 million Raspberry Pies have been sold before February 2015, making it the best-selling British com- puter. By November 2016 they had sold 11 million units, reaching 12.5m in March 2017, making it the third best-selling general purpose computer ever. A Raspberry Pi is a credit card-sized computer originally designed for education, inspired by the 1981 BBC Micro. Creator Eben Upton’s goal was to create a low-cost device that would improve programming skills and hardware understanding at the pre-university level. Due to its small size and accessible price, it was quickly adopted by tinker- ers, makers, and electronics enthusiasts for projects that require more than a basic micro controller (such as Arduino devices). The Raspberry Pi is slower than a mod- ern laptop or desktop but is still a complete Linux computer and can provide all the expected abilities that implies, at a low-power consumption level. The Raspberry Pi was designed for the Linux operating system, and many Linux distributions now have a version optimized for the Raspberry Pi. Two of the most popular options are Raspbian, which is based on the Debian operating system, and Pidora, which is based on the Fedora operating system. For beginners, either of these two work well; which one you choose to use is a matter of personal prefer- ence. A good practice might be to go with the one which most closely resembles an operating system you’re familiar with, in either a desktop or server environment. To experiment with multiple Linux distributions, NOOBS which stands for New Out of Box Software. There are, of course, lots of other choices. OpenELEC and RaspBMC are both operating system distributions based on Linux that are targeted towards us- ing the Raspberry Pi as a media center. There are also non-Linux systems, like RISC OS, which run on the Pi. Some enthusiasts have even used the Raspberry Pi to learn about operating systems by designing their own. 5.2 MACHINE LEARNING ALGORITHMS Data mining algorithms which carry out the assigning of objects into related classes are called classifiers. Classification algorithms include two main phases; in the first phase they try to find a model for the class attribute as a function Sahrdaya College of Engineering & Technology, Kodakara 22 SOCIAL NETWORK OF SAFE VEHICLES of other variables of the data sets, and in the second phase, they apply previously designed model on the new and unseen data sets for determining the related class of each record [1]. There are different methods for data classification such as Decision Trees (DT), Rule Based Methods, Logistic Regression (LogR), Linear Regression (LR), Naive Bayes (NB), Support Vector Machine (SVM), k-Nearest Neighbor (k- NN), Artificial Neural Networks (ANN), Linear Classifier (LC) and so forth. The comparison of the classifiers and using the most predictive classifier is very impor- tant. Each of the classification methods shows different efficacy and accuracy based on the kind of data sets. In addition, there are various evaluation metrics for com- paring the classification methods that each of them could be useful depending on the kind of the problem. We employed in our research three different machines learning algo- rithms.C4.4 decision tree , Support Vector Machine and Naive Bayes classifier. 5.2.1 C4.5 decision tree Classifier A decision tree (DT) is a flowchart-like tree structure, where each inter- nal node denotes a test on an attribute, each branch represents an outcome of the test, and each leaf node (or terminal node) holds a class label . The topmost node in a tree is the root node. During tree construction, attribute selection measures are used to select the attribute which best partitions the tuples into distinct classes. Three popular attribute selection measures are Information Gain, Gain Ratio, and Gini In- dex. When DTs are built, many of the branches may reflect noise or outliers in the training data. Tree pruning attempts to identify and remove such branches, with the goal of improving classification accuracy on unseen data. Decision tree learning uses a decision tree (as a predictive model) to go from observations about an item (represented in the branches) to conclusions about the item’s target value (represented in the leaves). It is one of the predictive mod- elling approaches used in statistics, data mining and machine learning. Tree models where the target variable can take a discrete set of values are called classification trees; in these tree structures, leaves represent class labels and branches represent conjunctions of features that lead to those class labels. Decision trees where the tar- get variable can take continuous values (typically real numbers) are called regression trees.Decision tree learning is a method commonly used in data mining. Sahrdaya College of Engineering & Technology, Kodakara 23 SOCIAL NETWORK OF SAFE VEHICLES The goal is to create a model that predicts the value of a target variable based on several input variables. An example is shown in the diagram at right. Each interior node corresponds to one of the input variables; there are edges to children for each of the possible values of that input variable. Each leaf represents a value of the target variable given the values of the input variables represented by the path from the root to the leaf. C4.5 is an algorithm used to generate a decision tree developed by Ross Quinlan. C4.5 is an extension of Quinlan’s earlier ID3 algorithm. The decision trees generated by C4.5 can be used for classification, and for this reason, C4.5 is often referred to as a statistical classifier. Authors of the Weka machine learning software described the C4.5 algorithm as ”a landmark decision tree program that is probably the machine learning workhorse most widely used in practice to date”. C4.5 classifier is a simple decision tree for classification. It creates a bi- nary tree to model the classification process. C4.5 builds decision trees from a set of training data in the same way as ID3, using the concept of information entropy. The training data is a set S = s1 , s2 , ... of already classified samples. Each sample si consists of a p-dimensional vector (x1,i , x2,i , ..., x p,i ), where the x j represent attribute values or features of the sample, as well as the class in which si falls.At each node of the tree, C4.5 chooses the attribute of the data that most effectively splits its set of samples into subsets enriched in one class or the other. The splitting criterion is the normalized information gain (difference in entropy). The attribute with the highest normalized information gain is chosen to make the decision. The C4.5 algorithm then recurses on the partitioned sublists. Once the tree is built, it is applied to each tuple in the dataset and leads to assign a class for that tuple [19][20]. While building a tree, C4.5 ignores the missing values. C4.5 allows classification via either decision trees or rules generated from them. Example of a decision tree in flowchart form is given in Fig.2.10 [26]. Algorithm 1 is the algorithm for C4.5 decision tree classi- fier model . Support vector machines (SVM) are supervised learning methods for classification and regression. Sahrdaya College of Engineering & Technology, Kodakara 24 SOCIAL NETWORK OF SAFE VEHICLES Fig. 5.1: Example of a decision tree in flowchart form. Algorithm 1 C4.5(D) Input: an attribute-valued dataset D 1: Tree = {} 2: if D ”is pure” OR other stopping criteria met then 3: Terminate 4: end if 5: for all attribute a ε D do 6: Compute information-theoretic criteria if we split on a 7: end for 8: abest = Best attribute according to above computed criteria 9: Tree = Create a decision node that tests abest in the root 10: Dv = Induced sub-datasets from D based on abest 11: for all Dv do 12: Treev = C4.5(Dv ) 13: Attach Treev to the corresponding branch of Tree 14: end for 5.2.2 SVM In machine learning, support vector machines (SVMs, also support vec- tor networks) are supervised learning models with associated learning algorithms that analyze data used for classification and regression analysis. Given a set of train- ing examples, each marked as belonging to one or the other of two categories, an Sahrdaya College of Engineering & Technology, Kodakara 25 SOCIAL NETWORK OF SAFE VEHICLES SVM training algorithm builds a model that assigns new examples to one category or the other, making it a non-probabilistic binary linear classifier (although methods such as Platt scaling exist to use SVM in a probabilistic classification setting). An SVM model is a representation of the examples as points in space, mapped so that the examples of the separate categories are divided by a clear gap that is as wide as possible. New examples are then mapped into that same space and predicted to be- long to a category based on which side of the gap they fall. In addition to performing linear classification, SVMs can efficiently perform a non-linear classification using what is called the kernel trick, implicitly mapping their inputs into high-dimensional feature spaces. A support vector machine (SVM) is an algorithm that uses a nonlinear mapping to transform the original training data into a higher dimension. Within this new dimension, it searches for the linear optimal separating hyperplane. A hyper- plane is a “decision boundary” separating the tuples of one class from another. With an appropriate nonlinear mapping to a sufficiently high dimension, data from two classes can always be separated by a hyperplane. The SVM finds this hyperplane using support vectors (“essential” training tuples) and margins (defined by the sup- port vectors). The basic idea behind support vector machine is illustrated with the example shown in Figure 1. In this example the data is assumed to be linearly separa- ble. Therefore, there exist a linear hyperplane (or decision boundary) that separates the points into two different classes. In the two-dimensional case, the hyperplane is simply a straight line. In principle, there are infinitely many hyperplanes that can separate the training data. Figure 1 shows two such hyperplanes, B1 and B2. Both hyperplanes can divide the training examples into their respective classes with- out committing any misclassification errors. Although the training time of even the fastest SVMs can be extremely slow, they are highly accurate, owing to their ability to model complex nonlinear decision boundaries. They are much less prone to over fitting than other methods. Sahrdaya College of Engineering & Technology, Kodakara 26