Showing papers by "College of Engineering, Pune published in 2016"

SVM classifier based grape leaf disease detection

Abstract: Grape constitutes one of the most widely grown fruit crops in the India. Productivity of grape decreases due to infections caused by various types of diseases on its fruit, stem and leaf. Leaf diseases are mainly caused by bacteria, fungi, virus etc. Diseases are a major factor limiting fruit production and diseases are often difficult to control. Without accurate disease diagnosis, proper control actions cannot be used at the appropriate time. Image Processing is one of the widely used technique is adopted for the plant leaf diseases detection and classification. This paper is intended to aid in the detection and classification leaf diseases of grape using SVM classification technique. First the diseased region is found using segmentation by K-means clustering, then both color and texture features are extracted. Finally classification technique is used to detect the type of leaf disease. The proposed system can successfully detect and classify the examined disease with accuracy of 88.89%.

A review of machine learning techniques using decision tree and support vector machine

Abstract: In this paper, the brief survey of data mining classification by using the machine learning techniques is presented. The machine learning techniques like decision tree and support vector machine play the important role in all the applications of artificial intelligence. Decision tree works efficiently with discrete data and SVM is capable of building the nonlinear boundaries among the classes. Both of these techniques have their own set of strengths which makes them suitable in almost all classification tasks.

Automatic detection of major lung diseases using Chest Radiographs and classification by feed-forward artificial neural network

Abstract: Chest Radiograph is the preliminary requirement for the identification of lung diseases. Tuberculosis; pneumonia and lung cancer these lung diseases are major health threat. According to recent survey; which was given by WHO; rate of people dying due to late diagnosis of lung diseases is in millions. Early diagnosis of these diseases can curb mortality rate. This paper proposes lung segmentation; lung feature extraction and it's classification using artificial neural network technique for the detection of lung diseases such as TB; lung cancer and pneumonia. We have used the simple image processing techniques like intensity based method and discontinuity based method to detect lung boundaries. Statistical and geometrical features are extracted. Image classification using feed forward and back propagation neural network to detect major lung diseases.

Multiphase flow analysis of hydrodynamic journal bearing using CFD coupled Fluid Structure Interaction considering cavitation

Abstract: In this study, a fully three-dimensional CFD analysis and multi-phase flow phenomena, has been successfully implemented for simulation of hydrodynamic journal bearing considering the realistic deformations of the bearing with Fluid Structure Interactions (FSI) along with cavitation. Mixture model is used to model cavitation in the bearing and parametric modelling is used for modifying the flow domain due to deformation. Both systems are coupled and design optimization based on multi objective genetic algorithm (MOGA), is used to obtain optimized solution of the attitude angle and eccentricity for the combination of operating speed and load. In the study of bearings with and without effects of cavitation, it is observed that maximum pressure values drop when cavitation is considered in the bearing. Also there is decrease in maximum pressure when elastic deformation in the bearing is considered. The oil vapour distribution goes on increasing with the increase in shaft speed, thus lowering the magnitude of the pressure build up in the bearing. Multiphase study of bearings with cavitation hence becomes extremely important in case of bearings operating with higher speeds. The experimental data obtained showed very good agreements with numerical results and considerable reduction in computation time is observed.

Integration of Smart Phone and IOT for development of smart public transportation system

Abstract: As population is burgeoning, there is an increase in the number of vehicles on the road and hence an upsurge in the problems associated with traffic management, especially the Public Transport. There is also an increase in the number of accidents and various other traffic related issues. Intelligent Transportation System (ITS) provides the solution to most of these problems by integrating existing technologies with the underlying infrastructure. With the advent of mobile technology and the ubiquitous cellular network, real time vehicle tracking for efficient transport management has become viable. The futile long wait for a bus to arrive can be avoided by Intelligent Public Transportation System. The omnipresence of Smart Phones and their ever increasing power at a very economical price makes them one of the most attractive options for developing IOT applications. Here, an approach based on the combination of technologies like GPS and Android is discussed which can assuage passengers who commute by the means of public transport. The user is furnished with explicit information about the current location of nearest buses approaching the bus-stop on a mobile application. Using readily available Android API's, technologies like 3G network and SMS based services in the existing mobile phones can reduce the cost and size of hardware required, as well as lead to a better output.

A survey on IoT applications, security challenges and counter measures

Abstract: Internet of Things (IoT) is a recent technology that permits the users to connect anywhere, anytime, anyplace and to anyone. In this paper, the various medical services of IoT such as Ambient Assisted Living (AAL), Internet of m-health, community healthcare, indirect emergency healthcare and embedded gateway configuration are surveyed. Further, the applications of IoT in sensing the glucose level, ECG monitoring, blood pressure monitoring, wheelchair management, medication management and rehabilitation system are analyzed. The analysis results show that the use of IoT in the medical field increases the quality of life, user experience, patient outcomes and real-time disease management. The introduction of medical IoT is not without security challenges. Hence, the security threats such as confidentiality, authentication, privacy, access control, trust, and policy enforcement are analyzed. The presence of these threats affect the performance of IoT, thus, the cryptographic algorithms like Advanced Encryption Standard (AES), Data Encryption Standard (DES) and Rivest-Shamir-Adleman (RSA) are used. The investigation on these techniques proves that the RSA provides better security than the AES and DES algorithms.

Identification of green building attributes for the development of an assessment tool: a case study in India

Abstract: Green building rating systems have been developed to measure the level of sustainability of buildings. Existing methods can be applied to different regions by addressing additional aspects such as varied climatic conditions and regional variations. This paper investigated the most widely used environmental building assessment methods, namely BREEAM, LEED, SB-Tool, CASBEE, LEED-India, GRIHA and Eco-housing. Comparative studies revealed that the existing assessment schemes had some limitations when applied to an Indian built environment. This necessitates the development of a new building environmental assessment scheme. An attempt is made to develop a framework to evaluate sustainability of buildings in India by applying principal component analysis (PCA). The PCA of 82 valid responses on the attributes measuring sustainability of buildings has extracted nine components: (1) site selection; (2) environment; (3) building resources and re-use; (4) building services and management; (5) innovative cons...

Secure file storage in cloud computing using hybrid cryptography algorithm

Abstract: Now a day's cloud computing is used in many areas like industry, military colleges etc to storing huge amount of data. We can retrieve data from cloud on request of user. To store data on cloud we have to face many issues. To provide the solution to these issues there are n number of ways. Cryptography and steganography techniques are more popular now a day's for data security. Use of a single algorithm is not effective for high level security to data in cloud computing. In this paper we have introduced new security mechanism using symmetric key cryptography algorithm and steganography. In this proposed system AES, blowfish, RC6 and BRA algorithms are used to provide block wise security to data. All algorithm key size is 128 bit. LSB steganography technique is introduced for key information security. Key information contains which part of file is encrypted using by which algorithm and key. File is splited into eight parts. Each and every part of file is encrypted using different algorithm. All parts of file are encrypted simultaneously with the help of multithreading technique. Data encryption Keys are inserted into cover image using LSB technique. Stego image is send to valid receiver using email. For file decryption purpose reverse process of encryption is applied.

Trust Management Model based on Fuzzy Approach for Ubiquitous Computing

Abstract: The current state of ubiquitous computing has been greatly influenced by emerging networking developments like Internet of Things (IoT), Future Internet etc. Adequate trust management is crucial to provide security. The entities involved in communication must be trusted for specific purposes depending on their role. Using trust model, devices can run trust computations and guide their behaviors. To this effect, a method is needed to evaluate the level of trust between devices. Trust models investigated so far discusses that devices face problems when communicating as transforming trust relationships from real to virtual world requires the negotiation of trust based on the security properties of devices. However, these models are developed in limited devices. This paper proposes a distributed trust model for device-to-device communication in ubiquitous computing. Mathematical model based on fuzzy rules to establish trust is presented. Fuzzy simulation of the model is presented to validate the findings. Simulation results show that proposed model calculates fuzzy trust values reliably.

Uncertainty-Estimation-Based Approach to Antilock Braking Systems

Abstract: Two new strategies based on uncertainty estimation are proposed for antilock braking systems (ABSs). In one of the strategies, the uncertainties and the disturbances are estimated using inertial delay control (IDC), and the estimates are used in a backstepping-control-based braking system. In the other strategy, in addition to uncertainties, the states are also estimated using an inertial delay observer (IDO) for a sliding-mode-control (SMC)-based braking system. No knowledge of uncertainties, disturbances, and the road adhesion friction coefficient or their bounds is assumed. The stability of the overall system is proven, and the schemes are validated by simulation and experimentation in the laboratory.

A boundary layer sliding mode control design for chatter reduction using uncertainty and disturbance estimator

Abstract: This paper proposes a new boundary layer sliding mode control design for chatter reduction. The control scheme uses a discontinuous control outside the boundary layer and switches over to uncertainty and disturbance estimator (UDE) based control inside. The problem of large initial control underlying the method of UDE, is also addressed with a modified sliding surface. The overall stability of the system is proved and the results are verified on an illustrative example and application to flexible joint system. The results show that the proposed method exhibits much better control performance than the baseline SMC using ‘sat’ function, for reduced chattering.

A novel training algorithm for convolutional neural network

Abstract: Many machine learning softwares are available which help the researchers to accomplish various tasks. These software packages have various conventional algorithms which perform well if the training and test data are independent and identically distributed. However, this might not be the case in the real world. The training data may not be available at one time. In the case of neural networks, the architecture has to be retrained with new data that are made available subsequently. In this paper, we present a novel training algorithm which can avoid complete retraining of any neural network architecture meant for visual pattern recognition. To show the utility of the algorithm, we have investigated the performance of convolutional neural network (CNN) architecture for a face recognition task under transfer learning. The proposed training algorithm may be used for enhancing the utility of machine learning software by providing researchers with an approach that can reduce the training time under transfer learning.

Vision-based approach for American Sign Language recognition using Edge Orientation Histogram

Abstract: Hand Gesture Recognition System (HGRS) for detection of American Sign Language (ASL) alphabets has become essential tool for specific end users (i.e. hearing and speech impaired) to interact with general users via computer system. ASL has been proved to be a powerful and conventional augmentative communication tool especially for specific users. ASL consists of 26 primary letters, of which 5 are vowels and 21 are consonants. Proposed Real-time static Alphabet American Sign Language Recognizer- (A-ASLR) is designed for the recognition of ASL alphabets into their translated version in text (i.e. A to Z). The architecture of A-ASLR system is fragmented into six consequent phases namely; image capturing, image pre-processing, region extraction, feature extraction, feature matching and pattern recognition. We have used Edge Orientation Histogram (EOH) in A-ASLR system. The system is developed for detection of ASL alphabets based on Vision-based approach. It works without using colored gloves or expensive sensory gloves on hand. Our A-ASLR system achieves the recognition rate of 88.26% within recognition time of 0.5 second in complex background with mixed lightning condition.

Solution to sustainable rural electrification in Myanmar

Abstract: About 70% population of Myanmar lives in rural areas where average electrification rate is mere 16%. Out of total 64,917 villages, about 57,557 villages are in remote areas, far away from the national grid. Myanmar is blessed with an abundance of energy potential and extensive renewable resources, including large amount of agricultural biomass waste. The rice husk is one of the major sources of biomass. The paddy rice production had been estimated at 28.9 million tons for 2014, producing 5.78 million tons of rice husks. Over 1000 rice mills across Myanmar are being powered by rice husk small scale biomass gasifiers. After 2001, few rice husk biomass power plant were installed by rural cooperatives/committees for rural electrification. The present investigation focuses to explore the self-sustaining energy service model to provide grid quality power to rural populations without the need of subsidies. The power system model focusing on both power requirements for the productive use and the neighboring villages/rural households at affordable tariff could become the most appropriate solution for the sustainable rural electrification. It is argued that the rice husk biomass power system installed and operated by rice millers is not only the sustainable and affordable option to rural electrification but also the financially viable business model to provide the grid quality power to rural population without grant or subsidy. Furthermore, the modern energy policy objectives-energy security, affordability, and sustainability are also met with the biomass rice husk power plant.

Vibration control of bridge subjected to multi-axle vehicle using multiple tuned mass friction dampers

Abstract: The effectiveness of tuned mass friction damper (TMFD) in reducing undesirable resonant response of the bridge subjected to multi-axle vehicular load is investigated. A Taiwan high-speed railway (THSR) bridge subjected to Japanese SKS (Salkesa) train load is considered. The bridge is idealized as a simply supported Euler–Bernoulli beam with uniform properties throughout the length of the bridge, and the train’s vehicular load is modeled as a series of moving forces. Simplified model of vehicle, bridge and TMFD system has been considered to derive coupled differential equations of motion which is solved numerically using the Newmark’s linear acceleration method. The critical train velocities at which the bridge undergoes resonant vibration are investigated. Response of the bridge is studied for three different arrangements of TMFD systems, namely, TMFD attached at mid-span of the bridge, multiple tuned mass friction dampers (MTMFD) system concentrated at mid-span of the bridge and MTMFD system with distributed TMFD units along the length of the bridge. The optimum parameters of each TMFD system are found out. It has been demonstrated that an optimized MTMFD system concentrated at mid-span of the bridge is more effective than an optimized TMFD at the same place with the same total mass and an optimized MTMFD system having TMFD units distributed along the length of the bridge. However, the distributed MTMFD system is more effective than an optimized TMFD system, provided that TMFD units of MTMFD system are distributed within certain limiting interval and the frequency of TMFD units is appropriately distributed.

An IoT framework for intelligent vehicle monitoring system

Abstract: In order to have safe vehicular traffic across the expressway a real-time monitoring has become a vital need for the today's intelligent traffic monitoring systems(ITS). In this paper we present a design and development of low cost and reliable Internet of things framework which consists of an array of RFID sensors for the real time tracking of the vehicle on its transit from one point to other point of the high speed expressway. The uniquely detecting capability of vehicle using RFID sensor network makes it a better choice compared to the image processing based systems. In this project a real-time stamps are taken from the array of RFID sensor network and the velocity of the vehicle is approximated in the real-time environment using Euler's algorithms. Here an Arduino platform with an Ethernet connection can be used as a core controller and the resultant data can be viewed on the internet using cloud computing.

Fusion of local and global features for classification of abnormality in mammograms

Abstract: Mammography is the most widely used tool for the early detection of breast cancer. Computer-based algorithms can be developed to improve diagnostic information in mammograms and assist the radiologist to improve diagnostic accuracy. In this paper, we propose a novel computer aided technique to classify abnormalities in mammograms using fusion of local and global features. The objective of this work is to test the effectiveness of combined use of local and global features in detecting abnormalities in mammograms. Local features used in the system are Chebyshev moments and Haralick’s gray level co-occurrence matrix based texture features. Global features used are Laws texture energy measures, Gabor based texture energy measures and fractal dimension. All types of abnormalities namely clusters of microcalcifications, circumscribed masses, spiculated masses, architectural distortions and ill-defined masses are considered. A support vector machine classifier is designed to classify the samples into abnormal and normal classes. It is observed that combined use of local and global features has improved classification accuracy from 88.75% to 93.17%.

Road traffic prediction and congestion control using Artificial Neural Networks

Abstract: In past decade, the problem of traffic has become severe due to industrialization especially in big cities. Hence, the urban population has to invest much valuable time during traveling.

Algorithm selection for classification problems

Abstract: A number of algorithms are available in the areas of data mining, machine learning and pattern recognition for solving the same kind of problem. But there is a little guidance for suggesting algorithm to use which gives best results for the problem at hand. This paper shows an approach for solving this problem using meta-learning. The paper uses three types of data characteristics. Simple, information theoretic, and statistical data characteristics are used. Results are generated using nine different algorithms on thirty eight benchmark datasets from UCI repository. The proposed approach uses K-nearest neighbor algorithm for suggesting the suitable algorithm. Classifier accuracy is taken as a basis for recommending the algorithm. By using meta-learning, accurate method can be recommended as per the given data, and cognitive overload for applying each method, comparing with other methods and then selecting the suitable method for use can be reduced. Thus it helps in adaptive learning methods. The experimentation shows that predicted accuracies are matching with the actual accuracies for more than 90 % of the benchmark datasets used. Thus it is concluded that the number of attributes, the number of instances, the number of classes, maximum probability of class and class entropy are playing a major role in classifier accuracy and algorithm selection for thirty eight datasets used for experimentation.

Investigation of DI-CI four-stroke VCR engine at different static injection timings using biofuel derived from non-edible oil source as a fuel

Abstract: This paper presents the effect of static fuel injection timings and blends of biodiesel with conventional diesel on the performance and emission characteristics of a DI-CI VCR engine. Blends of Honne oil methyl ester (HnOME) and diesel was used as fuel. The default value of static injection timing of the engine was 23° bTDC (before top dead centre). Injection timing was retarded and advanced from default value by 4° bTDC. Experiments were conducted at three levels of timings using the blends B20, B40, B60, B80 and B100 (pure HnOME). Conventional diesel was used as a reference fuel. The decrease in brake thermal efficiency for B20, B40, B60, B80 and HnOME compared to diesel at 19° bTDC were 5.4, 15.7, 13, 10 and 2.9% respectively. Brake thermal efficiency decreased by 4.2, 15.7, 13.5, 10.15 and 2.96% for B20, B40, B60, B80 and HnOME respectively compared to diesel at 27° bTDC. Nitric oxide emissions reduced for both advanced and retarded timings, but the reduction was more for retarded timing. Smok...

Chaos in discrete fractional difference equations

Abstract: Recently, the discrete fractional calculus (DFC) is receiving attention due to its potential applications in the mathematical modelling of real-world phenomena with memory effects. In the present paper, the chaotic behaviour of fractional difference equations for the tent map, Gauss map and 2x(mod 1) map are studied numerically. We analyse the chaotic behaviour of these fractional difference equations and compare them with their integer counterparts. It is observed that fractional difference equations for the Gauss and tent maps are more stable compared to their integer-order version.

Dual objective active suspension system based on a novel nonlinear disturbance compensator

Abstract: This paper proposes an active suspension system to fulfil the dual objective of improving ride comfort while trying to keep the suspension deflection within the limits of the rattle space. The scheme is based on a novel nonlinear disturbance compensator which employs a nonlinear function of the suspension deflection. The scheme is analysed and validated by simulation and experimentation on a laboratory setup. The performance is compared with a passive suspension system for a variety of road profiles.

Weighted Hybrid Decision Tree Model for Random Forest Classifier

Abstract: Random Forest is an ensemble, supervised machine learning algorithm. An ensemble generates many classifiers and combines their results by majority voting. Random forest uses decision tree as base classifier. In decision tree induction, an attribute split/evaluation measure is used to decide the best split at each node of the decision tree. The generalization error of a forest of tree classifiers depends on the strength of the individual trees in the forest and the correlation among them. The work presented in this paper is related to attribute split measures and is a two step process: first theoretical study of the five selected split measures is done and a comparison matrix is generated to understand pros and cons of each measure. These theoretical results are verified by performing empirical analysis. For empirical analysis, random forest is generated using each of the five selected split measures, chosen one at a time. i.e. random forest using information gain, random forest using gain ratio, etc. The next step is, based on this theoretical and empirical analysis, a new approach of hybrid decision tree model for random forest classifier is proposed. In this model, individual decision tree in Random Forest is generated using different split measures. This model is augmented by weighted voting based on the strength of individual tree. The new approach has shown notable increase in the accuracy of random forest.