Pusat Pengajian Kejuruteraaan Elektrik dan Elektronik - Tesis

Browse

Filter results by year or month

Now showing 1 - 20 of 200

Micro-Crack Detection Of Solar Cells Featuring Adaptive Anisotropic Diffusion Filter And Semi-Supervised Support Vector Learning
(0201-08)
Majid, Said Amirul Anwar B Ab Hamid @ Ab
In this thesis, a machine vision-based application for detecting micro-crack in an electro luminescence (el) image of solar cell is presented. The detection is a very challenging problem due to the complexity of the textural properties and background inhomogeneity of el images. Nevertheless, the micro-crack defect exhibits some unique properties such as high in gradient and low gray-levels. These properties together with the shape feature of the micro-crack are used in developing the detection algorithm. In this work, an image processing algorithm featuring an adaptive anisotropic diffusion filter and a segmentation technique based on twostage thresholding is proposed. The outcomes of this algorithm have demonstrated a highly accurate segmentation results compared to other standard methods. Based on the accuracy measure, the proposed methods achieve the highest f-measure of 0.0821. The local image features such as shape representation of the binary connected components are extracted and used in the machine learning to distinguish between cracked and good solar cells.
Identification Of Muscle Sound (MMG) In Human Body
(1999-04)
Khedher A. Hmood
From the biomedical engineering study on muscle fibers, there are two types of signals: Electromyography (EMG) signal and muscle sound (MMG). The EMG signal is generated by the human muscle and is detected by using stainless steel electrodes but the muscle sound MMG is detected using vibration detection sensor such as piezoelectric sensor. In this thesis, a suitable sensor has been used to detect the muscle sound in order to analyze its characteristics.
An Efficient Surface Mapping Technique Using Laser Triangulation For Image Recontruction
(2003-05)
Tjio Hok Hoo
An image acquisition system, which is used to capture 2-D images of object/scene for 3-D image reconstruction, is developed based on triangulation technique. The system consists of three devices, i.e. B/W CCO camera, laser diode and rotary table. These devices are placed in such a way that a triangle is formed. A grating line, which has a width of one pixel, is mounted in front of the laser diode. The purpose of the grating line is to project a stripe line on the surface of the object/scene. In this research, the positions of the camera and the laser diode are fixed. In order to acquire sufficient information for 3-D image reconstruction, the object/scene is rotated 360° about the azimuth of the rotary table. At every 15°, the rotary table is paused and a reflected image of the object is captured by the camera. Therefore, 24 different images are recorded in one complete rotation.
Electromagnetic Tomography For 2-D Mapping Of Moisture Content In Rice
(2003-10)
Lim, Meng Chun
This thesis deals with the development and application of electromagnetic tomography in food engineering applications, especially in moisture content profiling of cereals such as rice grain.
Development of behaviour-based mobile robot for goal seeking and obstacle navigation
(2009-03-01)
Oo, Cheng Hoe
In the field of robotic, there are some concerns that always grab the world attention which are the motion planning and control problem. Both of them will finally decide the collision-free trajectories for a robot in static or dynamic environments which may have many unknown obstacle around while seeking the goal. Thus this report describes the development of a behavior-based mobile robot for the obstacle navigation and goal seeking. The subsumption architecture will be utilized with the obstacle avoidance having higher priority over goal seeking. In fact, there are many possible obstacles that may keep the robot wandering around aimlessly seeking the goal. One of its main problems is local minimum problem which keep robot trapping in a corner. Owing to simplicity, wall following will be the better solution for that case. Beside that, a deadlock is also consideration in this paper. In the nutshell, the report will discuss all possible obstacles that may exist in the real world. However the goal seeking feature still has flaw when it is combined with the obstacle avoidance. Besides that, the designed robot will act as leader guiding the other robots which are the followers to avoid the obstacle around while searching the goal. This is called the leader-following.
Design and development of an intelligent battery charger station
(2009-03-01)
Teoh, Wee Wee
This thesis presents the design and development of an intelligent battery charger for simultaneous charging of four NiMH batteries. It is designed to satisfy the demands of high current and fast charge applications such as electrical vehicles which use batteries as the electrical source energy. The purpose of this design is to provide an efficient charging algorithm using a microcontroller PIC16F877 in order to protect against overcharging and reduce the recharging time. This design includes the discharging function. In this design, the control and supervision of the whole charging process is entrusted to a microcontroller, which able to find out the initial battery state , decide the fit way to charge the battery (in order to ensure a long cycle-life) and determine when the charge process must be finished. The proposed design provides multiple charging current options with automated selection of optimum charging rate for the battery being charged. The charging algorithm is programmed in the PIC16F877 by using PIC Basic Pro. There are four basic charging algorithm used in this charger: slow (10%C), soft (20%C), fast (55%C) and trickle (4%C). The experimental results obtained show that the charger is functioning properly with the state defined in the microcontroller. The current regulation is successfully done by using PWM control. The accurate termination of fast charging cycle and safe charging of batteries demonstrate the reliable functioning of the proposed design. The charging and discharging curve obtained is almost the same as stated in the theoretical characteristic. The charging time to charge the battery from empty to full is almost two hours. The battery voltage at the end of charging cycle obtained is around 1.38V/battery. The fast charge cycle bring the battery to approximately 90% of the full charge condition. The implementation of design’s objective is fulfilled and is supported by experimental results.
Artificial neural network for gas-oil flow pattern recognition using capacitance tomography data
(2009-04-01)
Tan, Kim Leng
The technique to recognize the oil and gas flow pattern in a pipe is needed in the oil and gas industry to monitoring the condition of the oil and gas pipe system. Any mistake or malfunction may lead to serious loss and endanger the workers and environments. Generally there are lots of flow pattern such as Empty, Full, Stratified, Bubble, Core and Annular. The Electrical Capacitance Tomography (ECT) technique is used to take the cross sectional data of the pipe. The Artificial Neural Networks (ANNs) is used to recognize the flow patterns. This project uses the Multilayer Perceptron (MLP) as the ANNs model. The MLP is trained, validated, and tested with the ECT data. The ECT data is divided into three groups, training, validation, and testing. The Matlab software is used to build the MLP architecture. The learning algorithms used for this project is the Levenberg-Marquardt algorithms and the Quasi-Newton algorithms. Result show that trained MLP is able to give a percentage of accuracy of 99.102% in oil and gas flow pattern recognition. This shows that the MLP is suitable to be used in the oil and gas flow pattern recognition.
Monolithic microwave integrated circuit (mmic) low noise amplifier
(2009-04-01)
Yip, Ching Wen
Monolithic microwave integrated circuit low noise amplifier (MMIC LNA) is project offered by Telekom Research & Development Sdn. Bhd (TMR&D).In this project, the LNAs was designed using the 0.15 μm GaAs pHEMT technology. The operating frequencies were 2.4GHz and 3.5GHz. LNA is an electronic amplifier that is required in receiver systems to increase the amplitude of the very low level signals from the antenna without adding too much noise. LNA is for wireless application. Software Advance Design System (ADS) was used to simulate the circuit and design the layout. LNA was designed using cascode topology with feedback techniques which produces better matching and unconditionally stable over the entire desired frequencies. For the 2.4GHz operation, the amplifier achieves gain of 14.949 dB, noise figure of 1.951dB and input reflection coefficient of -10.419 dB. With operating voltage supply at 3V, the total current consumption is 13 mA. Although the gain obtained is not within the specs 16dB – 23 dB, the value is acceptable. For cascade 3.5GHz, the amplifier achieves gain of 22.985 dB, noise figure of 1.964dB, input reflection coefficient of -12.427 dB and current consumption of 18 mA.
Hand tracking system as virtual input device
(2009-04-01)
New, Mung Yin
The objective of this project is to develop a vision-based hand tracking system, which can be used as an input device for communicating with the computer. In this system, the only external device required is a webcam. There are two webcams being used throughout the project, each for one hand. In this report, a hand tracking algorithm has been presented using Simulink® to track and recognize hand gestures for interacting with a computer. This algorithm is based on five steps; which are image acquisition, hand detection, hand segmentation, feature extraction, and gesture recognition. In hand detection, accurate skin colour model is being built, and is used for hand segmentation to segment the hand from the background. Feature extraction is performed to extract useful information from images. In gesture recognition, extracted information is used for recognizing the hand movement. Besides, a graphical user interface (GUI) is created for presenting the results. The results show that the system can successfully track the hand, and integration of the algorithm into useful applications such as motion controlling. From the project, it can also be seen that vision-based hand tracking system can be realized, and it provides a convenient way for human-computer interface.
Exploring graph cut technique for mammography images
(2009-04-01)
Ding, Nik Siong
Breast cancer is one of the most common diseases among woman nowadays. However, the presence of medical imaging tools such as mammography enables diagnosis of breast cancer to be conducted at beginning stage of breast cancer development. Mammogram is the image produced by mammography process. The presence of abnormal structures in mammograms will indicate that the woman is having a breast cancer. In this study, a new algorithm called graph cut was explored to evaluate its efficiency to detect the presence of abnormal structures in mammograms. Although graph cut technique is new but it has generated a lot interest among the researchers in the computer vision community. The primary reason for this rising popularity has been the successes of efficient graph cut algorithm in solving many low level vision problems such as image segmentation, object reconstruction, image restoration and disparity estimation. Masses and mircocalcifications are two most common abnormal structures in mammograms. In this study, a specific mammogram segmentation algorithm was developed based on the graph cut technique. This mammogram segmentation algorithm was used to test efficiency of the graph cut technique to segment the abnormal structures out of mammograms. The success rates of detecting these abnormal structures are high. For masses, graph cut technique able to detect the mass structures in all mammograms used to test this technique. Meanwhile, for microcalcifications, graph cut technique only can detect the microcalcification structures in mammograms which show low density breast. These segmentation results proved that graph cut technique is suitable to be used to detect the presence of abnormal structures in mammograms.
Harmonic effect on a three-phase induction motor performance
(2009-04-01)
Hj Ab Rashid, Mohd Hafizi
Three phase induction motor are by far the most widely used in industry, they popular in industry because they are more economical, last longer, and required less maintenance than other type of motor. Due to the increase application of drive motor like variable voltage drives, today's power systems are rich of harmonic content. A high harmonic content may affect the performance of the motor and could become a concern in the field. This research presents the effect of voltage distortion on three-phase induction motor performance with light load condition. In this research, three-phase full-wave controller (AC controller) acts as the harmonic source. A three-phase squirrel cage induction motor is tested by using different value of percentage 𝑇𝑇𝑇 𝑣 from 0% t0 10%. From this research, it founded that the torques for each harmonic is quite small compared to rated torque, it only 0.15% from its fundamental value. They are alternately negative and positive for the harmonics listed in ascending order, and the net torque is a negligibly small negative torque. When 𝑇𝑇𝑇𝑣 increase more than 5%, the torque motor drop more than 1% from the fundamental torque (for squirrel cage induction motor). The harmonic slip is close to one. The speed of induction motor is affected only about 0.07% from its fundamental speed when 𝑇𝑇𝑇𝑣 are more than 5%.
Development of a power saving system for underwater vehicle
(2009-04-01)
Yusoff, Mohd Ariff Redza
Nowadays, power uses age is very import in our modern world. Most of the power systems are using petroleum as the base. In this project, light will be the source of energy. In conventional ways, light will be converting to electrical energy using solar panel then transfer to load circuit using copper wire. This project transfer energy from ground to underwater, so it use fiber optic as a medium to transfer energy from ground to underwater. This project choose fiber optic because of two factor, first because it safe for underwater creature and the ability of the fiber optic itself. If the copper wire is use, it will transfer electrical current, if the is any leakage, it is dangerous for underwater creature near the wire or cable. It may shock by electric current. Light will be transfer in the fiber optic and will be convert to electrical energy in load circuit, if there is any leakage on the fiber optic, it still safe for the under water creature. The other import factor of fiber optic is it can carry unlimited intensity of light. This will reduce in cost and size of the system. This project also research on light multiplexing technique. Photodiode are use to convert light to electrical energy. The multiplexer light will energize the photodiode and develop electrical energy. This project is very safe for our environment and underwater creature.
Implementation of motion estimation and motion compensation using block matching algorithms for video coding
(2009-04-01)
Tan, Sin Peng
A motion estimation and compensation algorithm for video compression is implemented using MATLAB software. In this project, two types of block matching algorithm (BMA) have been developed that is the Exhaustive Search (ES) and Three Step Search (TSS). In this approach, the current frame of a video sequence is divided into a matrix of macro blocks that are then compared with corresponding block and its adjacent neighbors in the previous frame to find the motion vector that stipulates the movement of a macro block from one location to another in the previous frame. This movement calculated for all the macro blocks comprising a frame, constitutes the motion estimated in the current frame. A motion compensated image for the current frame is then created that is built of blocks of image from the previous frame. The matching of one macro block with another is based on the output of a cost function. The macro block that results in the least cost is the one that matches the closest to current block. From thesimulation results obtained, ES has a better PSNR performance compared with the TSS. The average search point per macro block for ES is also computed and the value is almost 9 times of that of the TSS. TSS technique proved to be the better BMA since it has a significantly smaller computation time with its PSNR performance almost the same as the ES.
Implementation of behaviour based mobile robot
(2009-04-01)
Tan, Tiong Cheng
Control architecture is an important investigative content in mobile robot. After studying the subsumption architecture and some other modified architecture, a hybrid architecture is introduced in the project, which is based on the subsumption architecture and integrates the behaviour based planning. Behaviour based mobile robot is claimed to enable task execution in unstructured and dynamic real-world environments. The process of behaviour based planning use the concept of fixed priority to coordinate the behaviours. It can hold the intention of sub-behaviours perfectly. Each of the behaviours switches smoothly between them. The simulation uses 5 layers control architecture including wander, avoiding obstacle, goal seeking and so on. The mobile robot solely depends on the information from the ultrasonic sensor for obstacle avoidance and LDR for goal seeking. The proposed algorithm is tested on a simple mechanical platform of mobile robot. The controller of the mobile robot is run on a “C” environment, which is Boost C. Besides the main objective of studying the behaviour based algorithm, multi agent robot system is another part of my project. A leader follower system is designed and realized. From the results, it shows this architecture is simple and effective for obstacle avoidance. It is robust and enables rapid response to the environment changes. The follower is able to keep track of the leader with the information received from the leader. As a conclusion, the proposed algorithm is effective in obstacle avoidance.
Detection and recognition of car number plate from video sequence
(2009-04-01)
Choo, Kar Soon
The goal of this project is to develop a system to localize or detect car number plate location and recognize the detected characters automatically. Car number plate recognition (CNPR) system is a challenging task due to the variety of car number plate format in Malaysia. A proper design has to be carried out in order to achieve high performance rate for this system. For this project, connected component analysis is used to detect or localize the car plate region. For character recognition part, K-Nearest Neighbors (KNN) classifier is designed to handle this task. The performance rate of the overall system is about 74% with all the car plate location is exactly detected, good segmentation of character and correct character recognition. Besides, this system is feasible to run in real time due to the simplicity of detection algorithm. A lot of works need to be done in the future in order to improve the performance rate of the system.
Image super resolution using interpolation
(2009-04-01)
Lee, Siang Hoe
Interpolation is a method to enlarge an image. It collects information from an image and uses this information to predict the unknown information in the upgrade size of image. The information in an image will affect the image resolution. Low resolution image has least information or smaller image size, and high resolution image has more information and bigger image size. In this project, a low resolution image is interpolated by four different interpolation methods, and this interpolated image is filtered by four filtering methods. These four interpolation and four filtering methods are nearest neighbor interpolation, bilinear interpolation, bicubic interpolation, Lanczos interpolation, smoothing filter, Gaussian filter, sharpening filter and unsharp masking filter. The final image of each method is compared, and the difference with ground truth image is measured. Based on the results, it shows that magnification ratio is inversely proportional to output image. As magnification ratio increase, the output image quality is decreased. In up sampling an image, interpolation methods should be considered first then only filtering methods. This because interpolation methods affects output result the most rather than filtering methods. Bicubic interpolation gives 18.5513 of MSE value and Gaussian filter gives 17.3515 of MSE value which both are the lowest MSE value among others interpolation and filtering methods for magnification factor of two.
Investigation on neural network input selection methods
(2009-04-01)
Lee, Hooi Khee
The purpose of this project is about the investigation of neural network input selection methods and its objectives is to eliminate the noise, irrelevant and redundant information that may improve the predictive power of an algorithm and to reduce the amount of data to process and the training time for effectiveness. Initially, a few input selection methods have been investigated and the best input selection method is chosen. The few set of methods include Principle Component Analysis (PCA) and stepwise selection strategies that assess usefulness of the inputs in the model are investigated. From the investigation, PCA is not suitable to be used in this project as this algorithm alter the original representation of the variables which may loss some of the important information. Although PCA is a powerful tool used for dimensionality reduction and feature extraction, this method is not optimal for dimensional reduction in target detection and classification applications. On the other hand, stepwise selection strategy is found out to be an efficient strategy to perform input selection task. Once the appropriate input variables have been selected, NN (neural network) will then be used to test this new training set for classification purpose. Then, the training and testing accuracy are evaluated. The best neural network has to be chosen correctly and efficiently in order to give the best training and testing performance to diagnosis results. Multilayer feedforward neural network (MLP) is chosen in this thesis to perform the clustering purpose due to the fact that MLPs are flexible, non-parametric modelling techniques and allowing us to perform any complex function mapping with arbitrarily desired accuracy. The MatlabR2008a, version 7.6, is the software that was used for the implementation in this project.
Ka band down converter
(2009-04-01)
Mohamed Sahid , Nur Arifah
This dissertation presents an approach of designing Ka Band Down Converter which are the combination of a Rat Race Mixer and Band-pass filter. Ka Band is the newest satellite broadcast band where Ka Band range is between 26.5 GHz to 40 GHz. The Ka Band Down Converter is designed to intend the operation at 38 GHz. The purpose of constructing the Ka Band Down Converter is to develop an application in wireless and radar systems. In order to develop the Ka Band Down Converter, the design construction is mainly focus on Rat Race mixer and Band-pass filter performances base on simulation perform using an Advanced Design System (ADS) from Agilent Technology. The simulation results showed the mixer and Band-pass filter design are able to operate at frequency 38 GHz for mixer and 233 MHz for filter. The Rogers RO4003C PCB type which is used in fabrication and the measurement reveal the hardware only able to perform up to 20 GHz only due to measurement equipment limitations.
Automation of voltage using analog voltage margining card
(2009-04-01)
Loh, Han Yang
In this project, voltage automation scripts for Analog Voltage Margining Card (AVMC) are being developed. The AVMC card is a card used to support analog validation of a target test system. It supports margining of target voltage supplies through voltage and simulated current controls. AVMC is used to replace ADM1066 chip for the voltage margining due to its accuracy and flexibility. Voltage automation scripts are written using bash language because it is suited for automation tasks. For the scripts to work, an operating system called System Validation Operating System (SVOS) such that user can margin the voltage by entering a single command. The voltage margining scripts generated is being tested on two types of Intel board which are Pinetrail and Tigerhill. Seven Pinetrail and five Tigerhill boards are used for testing the voltage margining scripts. When running a script on a board, the output voltage on the board is being measured using a digital multimeter. The measurement is then compared to the actual or desired value of the user. Voltage margining had been carried out to each available board and it shows that the scripts work. Hence, the voltage margining scripts for AVMC had been successfully developed and is currently used for running test purposes in Intel.
Gas pressure recognition in vacuum interrupter based on partial discharge using neural network
(2009-04-01)
Harianto, Buddy
The gas pressure of vacuum interrupter will be increased after 20-30 years in services. When it exceed 10 Pa, partial discharge may occur and lead to an interruption failure. Measures have to be taken to detect and recognize the phenomenon to avoid serious failure to the vacuum interrupter as well as the operation system. In this work, the gas pressure level recognition in vacuum interrupter is presented based on an artificial neural network framework. More specifically, this work used a multi-layer perceptron neural network for gas pressure recognition. All the input of the raw data comes from experimental works based on measurement of partial discharge light intensity using photomultiplier tube. In this experiment, the output of the raw data was generated based on different gas pressure level. Through this raw data, input feature extraction was done to reduce the training time required for the neural process model. A multilayer feedforward neural network with single hidden layer was used as the neural network architecture. The recognition was based on the different pattern generated between each gas pressure level. The network was trained in MATLAB software using batch training function. After a series of training, the model performance evaluation was carried out to determine the ‘error’. An optimum network configuration was determined the network that produced the minimal error respect to the training and testing. The recognition rate of the developed neural network was higher than the existing neural network system. As a result, the system shows the higher percentage of pattern recognition of 95% which is able to classify gas pressure level in the vacuum interrupter. .

Browse

Browsing Pusat Pengajian Kejuruteraaan Elektrik dan Elektronik - Tesis by Issue Date

Results Per Page

Sort Options