Study on different types of sensors and Machine Learning Algorithms used for Surveillance Automobile System

. A broad range of modern technologies is used for surveillance as well as security of passengers in automobiles. These technologies include face identification, fingerprint identification, Iris identification, and speaker recognition. Among these face recognition and fingerprint recognition are most commonly used for the surveillance of the vehicle. The demand for face identification in automobiles is increasing as it serves better than that fingerprint technology. The face identification system involves both face recognition and face detection. Researchers have tried to implement a face recognition system in automobiles to detect facial features from a captured image using different machine learning algorithms. This will prevent an unauthorized person to enter the car and will lock the ignition system until and unless the person is recognized. This paper gives a brief review on different types of sensors combined with processing unit and different types of machine learning algorithms used for the vehicle security and tracking system using different technologies.


Introduction
Over the decades there has been a vast technological progression in the automobile sector.There have been various measures to monitor speed, distance, seat belt pre-tensioner, door locks, airbags, hand breaks, and keyless entry systems.Keyless entry authentication is still a major concern for the automobile industry.As the person driving can be a authenticate driver (or) a thief trying to steal the car.Various advancements have been made to resolve this problem that include fingerprint identification and face recognition system.Face identification is widely used as it includes touchless technology.Face identification includes both face recognition and face detection.Face detection helps to detect a human face in a proper image and face recognition is to recognize whether a particular face matches the stored database.Face detection is a phase where the program detects whether an image has a face or not.This uses some classifiers named Eigenface Classifier, Fisher face Classifier, Haar Classifier, and LBP Classifier.The vehicle tracking can be done using GSM and GPS technology.The Global Positioning System is used to determine the current location of the vehicle along with the latitude, longitude and altitude.The Global System for Mobile Communication serves as a communication system between the vehicle and the authorized persons.The integration of vehicle security system as well as tracking system prevent antitheft activities and alerts the person in case of an emergency.This review paper mainly focuses on different types of biometric systems used in the vehicles for the authentication purpose and vehicle tracking system using different sensors.

A review on Vehicle Security and Tracking System
The paper [1] describes about a alcohol detection system which includes MQ3 alcohol sensor, used to detect the levels of alcohol in humans.This gadget detects the concentration of alcohol in the air, and if the concentration of alcohol exceeds a certain level, the ignition system of the automobile is turned off, and the vehicle will not start.The system comprises of 3 parts i.e., sensor part, processing part, and display part.Sensor part is used to sense alcohol concentration in air and send the same in form of voltage signals to processing part.Processing part receives voltage signals and converts them to analog signals that can be used as a standard for determining alcohol concentration.Display part receives the processed signal and displays the data in the form of diagrams to users (or) drivers.If the alcohol content exceeds the sensors capacity it displays 'detected' with the amount of alcohol content and if it's less than that it displays 'no alcohol detected'.
The paper [2] demonstrates a face detection system that is used to recognize the driver's face and compare it with the predefined face which is stored in the PC memory unit.The security system includes a memory card module to store pictures of the owners and their families.As the driver drives the vehicle the image of his/her face is captured and compared with the stored images.Face detection algorithm includes three steps: (a) Detect a face to track, (b) Identify facial features to track, (c) Track the face.In order to track a face, the first step is to detect it.Here the Cascade Object Detector is used to determine the location of a face with in a frame.The detector uses Viola-Jones algorithm for the detection purpose.The second step is to identify facial feature that will allow to track the face.The facial feature considered here is skin tone.The Histogram based tracker is used for tracking which employs CAM-Shift algorithm to track the object by measuring the histogram of pixel values.Once the face matches with the predefined faces the driver can use the vehicle.If in case it doesn't match, the face of the person is captured and is send as a MMS to owner and nearby police station to track the vehicle using GPS and GSM module.
The paper [3] describes about a face Recognition Car ignition System where the key is replaced with a user's face to start the ignition of the engine.The facial recognition system is carried out by combining the face detection and face tracking system algorithm found in MATLAB with the help of Raspberry pi B. The prototype used here consists of Micro controller and GSM service.To achieve high detection accuracy, primarily appearance-based methods are used.Face colour and eye is an important feature for face detection and recognition process.HAAR Classifier is used to detect the face of the person from a video sequence.Adaboost algorithm is used to detect whether the image is a face or not.Principal Component Analysis is used to compare the principal components of the face with the predefined face from facial database so that the authenticated face is determined.The hidden camera placed in the car is used to take the picture of the person whosoever sits in the driver's seat.With the help of the GSM framework the owner of the vehicle is informed about the burglary and the robbery image is captured and saved in the database for police verification.After the owner's authentication, the car is started immediately with ignition.The sensors of the black box provide adequate data to the system, and the owner is provided with accurate information regarding the status of the car.
The paper [4] describes the use of machine learning algorithms like the haar cascade algorithm and local binary pattern histogram for facial recognition in automobiles.The sensors attached to the vehicles are used for detecting human intervention around a considerable region and the algorithm helps to recognize the right person or the owner and gives the access for ignition which in turn supplies power.The haar cascade algorithm is used to identify a particular object from the images and extract haar like features from it.In order to obtain Haar-like features, the sum of pixels under white and black areas is subtracted.The local binary pattern histogram is a face recognition algorithm.It works by thresholding a region of pixels around each pixel and then converting the result into a binary value, which is then visualized as a histogram.When a person is authorized or recognised the power flows from the battery to traction motors for propulsion.In case an unauthorised person tries take a control of the vehicle the access is denied.A comparison of local binary pattern histogram algorithm, eigen faces algorithm and fisher face algorithm is made from which local binary pattern algorithm is considered the best for detecting and recognizing faces.
The paper [5] explains a proposed System.The proposed system uses a GPS/GSM SIM900A module that combines GPS and GSM functionality.GPS data provides the current location of a vehicle, while GSM technology is used to send an alert message to the vehicle's owner.The GSM modem accepts a SIM card i.e., SIM900A.The system comprises of a LPC2148 microcontroller, several types of sensors, PWM, GPS, and GSM.The system includes a gas leakage monitoring system utilizing the MQ6 sensor and a temperature sensor DS18B20 for the safety of the traveller.LM35 sensor and gas leak detection sensor is used to prevent overheating of the engine and gas leakage.The authorized person will receive an alert message if one of the sensors is activated.The Hall Sensor is incorporated into the system to measure the speed of the vehicle.The owner receives an alert message from the GSM modem when the vehicle deviates from the predefined path and when it exceeds the speed limit.The vehicle can be stopped using commands.
The paper [6] proposes a framework that combines a face detection subsystem with a GPS (Global Positioning System) module, a GSM (Global System for Mobile Communications) module, and a control platform for an embedded smart car security system.This subsystem uses an optimized AdaBoost algorithm to detect faces in cars when nobody should be in the vehicle and can make an alarm either inaudibly or loudly.Other modules help users keep track of cars regardless of their location, even if they are lost.This system prototype is built on the base of one embedded platform in which one SoC named "SEP4020" (works at 100MHz) controls all the processes.Interestingly, this car security system has proven to be a viable solution, and it is also much cheaper and 'smarter' than other methods.
The paper [7] explains about design and development of a vehicle tracking system that gives necessary information about the vehicle.The controllable system designed has a combination of a GPS receiver module and a GSM communication network.GPS provides the location data, which are then transferred via GSM modem which acts as a communication tool using SMS (Short Message Service).Here GSM-GTM 900 is used for communication purpose and microcontroller MSP430F149 is utilized to process out information.The system includes three main components which include SMS receiver, GPS receiver with data processing unit, and SMS transmitter modules.The installed system works by activating the engine without a key which sends a signal to the microcontroller.The microcontroller detects the signal received and notifies the owner via mobile phone and immediately stops the engine by disconnecting its power supply.This helps the owner to have control over the vehicle.The extra features include DND activation (Do Not Drive) to confuse thief, voice announcement system in public.The paper [8] describes the different components integrated into a system for the surveillance of the car.The different components consist of a Raspberry-Pi mini-CPU processing unit, camera, GSM module, PIR sensors, camera module, and buzzer.PIR (Pyroelectric Infrared Radiation Sensors) attached to the integral system help to detect the intrusion.This in turn triggers the buzzer.The GSM module interacts with the raspberry pi and the camera module attached captures the image and sends it to the GSM module which later converts it to MMS and sends it to the owner's phone.At the same time, an email is sent to the proprietor with the user's image for authentication.Raspberry PI Python 2.7.5 integrated design environment was used to code the program.The car can be immobilized in case of theft by a single-word text message through the owner's phone.
The paper [9] illustrates the construction of an authentication system for the security of rental cars.The algorithms used for image processing are Advanced Haar Cascade and Adaboost algorithm.The authorized drivers of the rental companies can only drive the car with permission from the owner.In case of an accident, it tracks the location of the car with other additional details and sends a message to the owner with help of GPS and GSM modules in the built-in system.This is achieved through an android studio app on mobile phones.The infrared sensor is attached near the driver's seat to detect the motion, and the raspberry pi camera fixed inside the car's mirror captures an image and sends it to the system for further processing.The Haar cascade algorithm detects the face in the image and eliminates other irrelevant features.Unknown faces can be discarded easily in this way.
The paper [10] describes a vehicle security system that focuses on face recognition, helmet detection, and alcohol detection.The procedure initiates with helmet detection where IR sensor is used to determine whether the rider has worn a helmet or not.The authentication proceeds further when it is detected.This helps to turn on the engine.The next step is to identify whether a person has consumed alcohol or not.If any trace of alcohol is found the MQ3 sensor senses it and ignition is turned off.Final step is to detect human faces and recognise it with application of haar cascade algorithm.Ignition turns on if face matches database face and incase it doesnot match the owner gets notified about this.ADXL sensor is used to sense accident and give an alert to the persons whose details are saved in the database.
The paper [11] explains the application and hardware system used to protect the vehicle from theft.The hardware components include a Node MCU microcontroller, a relay switch, a GSM module, and a GPS module.Vehicles equipped with these hardware components are connected and controlled via mobile application.The app allows user to register their own information as well as multiple vehicles information and keep a track of their location as they travel.All these data are stored using MQTT cloud service in Node MCU microcontroller.The ignition is linked with the relay switch and Node MCU microcontroller.Thus, the application serves as a medium to operate the vehicle ignition system when required i.e. (to on (or) off ignition system).This feature secures the vehicle.
The paper [12] presents an antitheft system that uses a password procedure to turn on the engine.The IR sensor is used to detect the key in the keyhole.This sends a signal to the microcontroller.The microcontroller used here is ArduinoMega2560.The engine gets turned on if the user enters the correct password and in case the user enters an incorrect password more than three times the relay which is set in between the microcontroller and engine goes to open condition.The owner gets a message that 'someone is trying to access your vehicle' and an alert message is sent to a nearby police station tracking the vehicle's location.Here the owner gets an advantage to stop the vehicle by sending a 'STOP' message to the SIM card integrated into the system.The additional features include emergency calling through Sim900a GSM Module and a vehicle accessing feature in an emergency using a Hall effect sensor to allow only 25-wheel rotations.The paper [13] illustrates the antitheft and tracking mechanism for vehicles.The mechanism includes seven components -RFID Reader, RFID tag, Microcontroller, Vibration sensor, GSM, GPS, and LCD Display.The RFID Reader and tag play an important role in the detection of theft.The RFID tag is embedded in the key, and the RFID Reader is mounted inside the car.The RFID tag and Reader both have a unique code.The vehicle gets started only if the unique code present in both the RFID Reader and tag matches.If the unique code present in both doesn't match the user (or) owner receives an alert message using the GSM module.The vibration sensor positioned in the vehicle senses the vibration when both the RFID reader and tag fail to match.This gives an intimation about the vehicle to the user and keeps him (or) her updated about the current location of the vehicle.This aids to find the lost vehicle.

Table 1 .
Components and sensors integrated into the system to detect the vehicle.

Table 2 .
Algorithms used for face Recognition.