M.Sc. Engg.

Browse

Search Results

Now showing 1 - 10 of 161

A Technique for Assuring Secrecy and Lossless Properties of Digital Image
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2020-09) Tanveer, Md. Siddiqur Rahman; Alam, Prof. Dr. Kazi Md. Rokibul
Due to various disease diagnosis, the volume of medical data is rising fast. Also, for telemedicine, while medical image transmits over the public network, the distortion of pixels may cause erroneous disease diagnosis. Here, encryption of the image by multiple chaos-based schemes along with DNA cryptography can be a safeguard. As chaotic schemes are very sensitive to the initial conditions, a small difference in the initial conditions yields entirely uncorrelated sequences that assure the strength of encryption. To get high randomness, several DNA encoding and computing rules are deployed. This thesis proposes a multi-stage chaotic encryption technique for the medical image through Logistic map along with Lorenz attractor and DNA cryptography, where both schemes possess the most significant value of control parameters. Thus, their consecutive deployment generates colossal chaotic sequences that ensure the robustness of the proposed technique. At first, the usage of the Logistic map with SHA-256 hash value generates a chaotic sequence that converts the plain medical image into a confusing image. Now, this sequence is used to create a confusion key to encrypt this blur image. Later on, to overcome the limitations of DNA computing rules and to get high randomness, encode this blur image and Lorenz attractor based key according to DNA encoding rules. These rules are determined randomly from eight encoding rules. Then, execute DNA operations between encoded blur image and Lorenz key using the four DNA computing rules and these rules are also determined by chaotic logistic sequence. Thus, the ultimate cipher is generated. Then, to approve the potency of the cipher, a randomness test according to NIST, security and statistical analyses and comparisons are performed.
Automatic Detection and Classification of Diabetic Lesions for Grading Diabetic Retinopathy Using Fuzzy Rule-Based Classification System
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2020-10) Afrin, Rubya; Shill, Prof. Dr. Pintu Chandra
Diabetic Retinopathy (DR) is a chronic, progressive retinal disease which is the most common cause of legal blindness. The disease is threatening to eyes as it shows no signs of visual abnormality at the initial stage. It gradually decreases patient’s eye-sight and drives into blindness in future. Hence, the early detection of DR is vital to prevent the complete vision loss of diabetes patients. Traditional diagnosing system of DR requires quite trained ophthalmologists for monitoring the retina periodically. Moreover, several physical tests like fluorescein angiography, visual acuity test, and ocular coherence tomography are involved to detect DR which also require a lot of time to process. In this paper, a fuzzy rule-based classification technique is proposed for automatic detection and classification of retinal lesions for grading DR. The proposed technique consists of preprocessing of fundus images, extraction of candidate retinal lesions, formulation of feature set, and classification of DR. In the preprocessing phase, the technique eliminates background noises and extracts optic disc from the retinal fundus image. Four leading lesions; blood vessels, microaneurysms, haemorrhages, and exudates are extracted using different image processing techniques and two textural features; contrast and homogeneity are calculated in the detection phase. Then, these six input features; blood vessels area, microaneurysms count, haemorrhages area, exudates area, contrast and homogeneity are fed to fuzzy if-then rule-based classifier for predicting normal, mild NPDR, moderate NPDR, severe NPDR and PDR stages of DR.. A total of 520 retinal fundus images are collected from four public database; STARE, DIARETDB0, DIARETDB1 and MESSIDOR and the images are successfully classified by the fuzzy rule-based classifier with accuracy up to 92.42%. The sensitivity and specificity of the classifier are 92.44% and 94.29% respectively. The simulation result on publicly standard image datasets exhibits that the intended technique gives promising results in identifying retinal lesions and it has better capability of classifying several stages of DR compared with other existing automatic diagnosing system.
A Study on the Effect of Seam, Color and Boundary Priors on Salient Region Detection
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2019-12) Islam, Aminul; Ahsan, Prof. Dr. Sk. Md. Masudul
Being the best creation of almighty Allah, human beings have the ability to analyse a visual scenario. They not only see an image, but can judge the importance of different parts of a visual area. They can easily differentiate a running car from its background, can tell the color of different objects in an image and can focus attention to some important parts of a visible scene. The value of this visual power of human is easily understood if one thinks about these. Achieving this wonderful human quality using machine is one of the most precious goals of today’s scientific researches. With the continuous advancement of imaging technologies, more and more visual data are being collected all over the world. But, a major portion of these data are left unprocessed. Image processing is used to process these types of visual data. For all image processing techniques, the initial goal is to ﬁnd some target region for extracting information from the image. Saliency detection is the technique of computationally finding important regions of an image. It is usually done using the contrast information present in an image. Seam map is the combination of cumulative summation of energy values from different directions. In this thesis, a combined method is proposed which uses seam importance map along with boundary aware color importance map. Color importance map is the weighted average of different color channels of Lab color space. Some intermediate combinations which are closer to the proposed optimized version but differ in the optimization technique are also presented in this thesis. Several standard benchmark datasets including the famous MSRA 10k and ECSSD datasets are used to evaluate performance of the suggested method. The proposed saliency model has been compared with several state of the art methods for each dataset. The qualitative and quantitative results from those omparisons make things easier to understand. Besides that, comparison with those state of the art techniques and precision recall curves and F-beta values found from the experiments on several datasets prove the superiority of the proposed method. This robust saliency model can improve almost all types of vision related applications including object detection, robotics, medical image analysis, content aware image resizing etc. In this research, the proposed architecture of saliency detection technique has been applied into a simple implementation of pedestrian detection. Its application has signiﬁcantly improved the result of pedestrian detection in terms of performance and accuracy.
Design and Demonstration of Smartphone-Based Colorimeter
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2019-12) Rani, Saptami; Islam, Prof. Dr. Md. Rafiqul
In this thesis work, smartphone-based colorimeter is designed and practically implemented utilizing the in-built sensors like CMOS camera, flash LED, and high-power processor of the smartphone. The developed totally self-contained colorimeter is low-cost, light-weight, robust, field-portable and easily accessible. It has smart sensing facilities without the requirement of additional optics and external power supply. The device can be applied for real-time and on-site measurements of different types of analytes in the fields of environmental research, biomedical applications, and agriculture, which are completely absent in the currently used conventional bench-top type colorimetric instruments. In the real-world, for most of the colorimetric detection, attributes of color such as wavelength, intensity, saturation, etc. vary simultaneously according to the variation of analytes. The conventional smartphone-based colorimeters are mainly designed to measure the analytes considering the change in color information in only one domain which limits the colorimetric measurement in some specific analytes with a narrow band of detection. In this research, the developed smartphone-based colorimeter can quantify any analytes through multiple nonlinear regression based colorimetric assessment in a wide range of detection considering the variation of color attributes in all significant domains. To demonstrate the smartphone-based colorimeter a 3D optical enclosure is designed and fabricated for ensuring the constant illumination and hence to improve the SNR by isolating the measuring platform from the environmental illumination. Self-referencing is a unique characteristic of the instrument to calculate the color ratio with respect to the colorimetric information of the sample. A customized Android-based smartphone app is developed for the complete functioning of the developed colorimeter. The app is developed with the graphical user interfaces of calibration, assessment of the real-time or previously recorded test samples, save, and share the results of colorimetric measurement for multiple analytes of different colorimetric tests. For the first time, a novel wavelength estimation algorithm is developed to estimate the wavelength information of the reflected light of colorimetric measurement. To justify the performance of the developed colorimeter, three different colorimetric tests are demonstrated in this research named as Rhodamine B concentration quantifier, digital pH meter, and chlorine concentration quantifier using the Xiaomi Redmi Note 4 smartphone. Three different colorimetric characteristics are found for the three samples: only color tone changes significantly with the variation of Rhodamine concentration, the wavelength of color varies significantly with the variation of pH value in water, and color intensity, wavelength, and saturation all vary simultaneously with the variation of chlorine concentration. For all of the three colorimetric tests, the performance of the designed smartphone-based colorimeter is found excellent compared to the conventional colorimeters. The average error of RhB concentration quantifier within the detection range of (0.2-4.0) PPM is 0.95% whereas the chlorine concentration quantifier shows an average error of 1.16% for the detection range of (0.1-8.0) PPM with sensitivity 0.1 PPM. On the other hand, the digital pH meter detects pH value in the range of (4.0-9.0) with an average of 0.0876% detection error. It is noted that the present smartphone-based colorimeter is designed and demonstrated using three analytes but the developed device can be applied to measure any colorimetric analytes by proper calibration using the developed smartphone app. So, the developed martphonebased colorimeter could be a cost effective common platform for the colorimetric measurement of various analytes in different fields of applications.
Development of Surface Plasmon Resonance Sensor for Detecting Food Preservatives
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2020-02) Moznuzzaman, Md.; Islam, Prof. Dr. Md. Rafiqul
Food preservatives and adulteration is the universal concern of recent days. Particularly, developing states are the main maltreated with food contamination and it is a thoughtful problem in recent years. Formalin is a chemical compound that commonly present in food used for the preservation. Its frequent and illegal addition with food is a danger for human health and psychology. The recurrent ingesting of formalin contaminated food causes uncompromising health sicknesses. This critical issue causes fatal disease like chronic cancer. Therefore, identification of formalin in food is an extreme need, which is becoming a general problem in emerging countries. In this dissertation, formalin is detected quantitatively by designing a Graphene-MoS2 amalgamated 2D nano sheets with a TiO2-SiO2 nano-layered surface plasmon resonance (SPR) based sensor. This sensor distinguishes the presence of formalin utilizing the attenuated total reflection (ATR) approach and inspecting the reflectance vs SPR angle and transmittance vs surface plasmon resonance frequency (SPRF) attributes. Analytical approach for analyzing the sensor performance parameters has been carried out using MATLAB commercial software. The quantitative effect analysis of individual and amalgamated Graphene-MoS2 with TiO2-SiO2 layers has been studied for sensitivity, detection accuracy and quality factor. In addition, optimization of Silver layer thickness is carried out for sensitivity, detection accuracy and quality factor individually. Electric field distribution through the sensor has been investigated and analyzed by using YEE algorithm on the Lumerical FDTD solution commercial software. An alternative composite layer sensor structure has also been designed and developed in a configuration of Graphene-PtSe2-Ag-ZnO with BK7 glass prism. Formalin is detected successfully using this sensor structure by angular investigation method. The performance of this alternative sensor structure for formalin detection has been analysis analytically and it shows an outperform with very high sensitivity. A comparative study among the performance of different composite sensor structure and the proposed sensors are also presented. Another comparison has been carried out between the existing sensor structures and the proposed sensor structures. This is the silver integrated composite sensor that shows highest performance reported by SPR technology. Finally, recommendations for additional research has been anticipated.
Protein Folding Optimization in a Hydrophobic-Polar Model for Predicting Tertiary Structure Using Fruit Fly Optimization Algorithm
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2020-02) Chatterjee, Sajib; Shill, Prof. Dr. Pintu Chandra
The prediction of the three-dimensional structure of a protein from its amino acid sequence is an experiment that is very much well known optimization problem which is known as the Protein Folding Optimization (PFO) in many years. The PFO problem states to the computational problem of how to predict the local structure of a protein from its amino acid. PFO problem is the NP-hard and most challenging problem. Various kind of optimization algorithm already applied for solving the PFO problem, but none of the existing algorithm not provide the accurate result within optimal time. Fruit Fly Optimization Algorithm (FOA) is a recent metaheuristics algorithm that have the intensity and diversity characteristics of searching technique. Therefore, we applied FOA for solving PFO problem in the HP (Hydrophobic-Polar) cubic lattice model. In order to increase the convergence of the FOA, we have designed and developed three different operators of FOA: smell-based search, local vision-based search and global vision-based search technique for the perspective of PFO problem. The proposed algorithm is based on two extra mechanisms centroid hydrophobic and moderator mechanism, which are accountable for improving the accomplishment of the algorithm. The centroid hydrophobic mechanism tries to move the hydrophobic monomers to the center position of the structure. The moderator mechanisms try to move a part of monomers in the protein sequence each possible directions and place at the position where the maximum energy value found. This two extra mechanisms improved the performance of the propose algorithm magically. Moreover, we have developed a reconstruction operator for producing an accurate 3D structure of protein sequences by erasing overlapping in cubic lattice points. The experiment result shows of our proposed Fruit Fly Optimization Algorithm for Protein Folding Optimization (PFO_FOA) provide better accuracy than the existing algorithms.
Microstrip Low Pass Filter Realized By EBGS Assisted T-Line
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2019-06) Amanullah, Mohammad; Mollah, Prof. Dr. Md. Nurunnabi
The EBGSs provide passband and stopband performance. In this project the planar conventional structures and the defected ground structures have been described. In uniform conventional structures (circular, rectangular and triangular pattern) it is observed in the scattering parameter performance that the number and the depth of the ripple (-5 db) of the passband are high and the stopband is not much wider. Maximum 10 dB RL bandwidth is 7.4 GHz, the 20 dB rejection bandwidth is 4.8 GHz and the center frequency is 10GHz. The maximum value of isolation is found to be 54dB. Then, the structures with different filling factor (FF), different number of element and same area of different structures are investigated. Then, the results of non-uniform conventional are observed and the results are ripple free in passband. The 10 dB passband RL BW is found to be 6.88 GHz and the 20 dB rejection bandwidth is not found with maximum isolation of 13.26 dB. Next the uniform dumbbell shaped DGS is observed. The 20 dB rejection bandwidth is 4 GHz. The maximum value of isolation is found to be 62 dB. The 3dB cutoff frequency is shifted to 4GHz which provides more than 200% compactness from the conventional structure. Next the result of non-uniform dumbbell shaped DGS have been observed and 10 dB RL-BW is 4.4 GHz, 3 dB cut-off frequency is 4.2 GHz, 20 dB IL-BW is 17.5 GHz and maximum peak of IL is -55 dB have been found. Finally, the hybrid dumbbell shaped DGS is observed and we have found the best performance. The best performance of 10 dB RL-BW is 4.1 GHz, 3 dB cut-off frequency is 4.4 GHz, 20 dB IL-BW is more than 19 GHz and maximum peak of IL is -68 dB. So, we have found the best performance in the hybrid dumbbell shaped DGSs. The designs are more compact than the conventional designs and able to remove the unwanted spurious transmission in the band rejection region. Finally, the insertion loss (IL) and return loss (RL) performances of the LPFs (non-uniform triangular dumbbell shaped DGSs and hybrid DGSs) have been compared. It is seen that they provide improved performance in terms of the ripple suppression in the passband and IL bandwidths. Thus we have realized microwave LPF by EBGS assisted T-Lines. At last the physical representation of designs and the validification of the results are provided as for the validification of the work.
Finite-State Predictive Current Control of a Simplified Three-Level Neutral-Point Clamped Inverter
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2019-11) Tariquzzaman, Md.; Habibullah, Dr. Md.
Multilevel inverter is one of the most important parts in renewable energy based power generating section as well as in motor drive applications. The quality of an inverter system depends on current total harmonic distortion (THD), switching loss, fault tolerant ability, dynamic responses, voltage stress, common mode voltage etc. Multilevel inverter yields low current THD, less voltage stress across the semiconductor switches and low switching frequency and thus less switching loss. However, using more number of semiconductor devices and neutral point voltage variation are the common problems for a neutral point clamped inverter. This is why different topologies of multilevel inverter are available in the literature in order to solve the aforementioned problems. The control scheme of a multilevel inverter also plays an important role to guarantee system’s performance. Recently, model predictive control (MPC) draws much attention to the researchers for its intuitive features and easy handling of nonlinearities of a system. The controller uses system model to predict the future behavior of the system over a prediction horizon. The control objectives are met by minimizing a predefined cost function that represents the expected behavior of the system. The objective of the proposed research work is to control the output load current of a three level simplified NPC (3L-SNPC) inverter topology using MPC. The simplified NPC inverter is considered, because less number of semiconductor devices used in the topology, even though further investigation is required on different factors such as voltage stress, common mode voltage, losses and switching frequency. MPC is used as controller because it can handle the dc link capacitors voltages balancing problem in a very intuitive way. Moreover, the average switching frequency reduction and over current protection can be easily implemented. Simulation results show that the proposed 3L-SNPC yields similar current THD, transient and steady state responses, voltage stress on the switches at the load side and over current protection capability as the conventional diode clamped based NPC inverter system. The two dc-link capacitor voltages are balanced properly with a neutral point voltage variation of close to zero. However, in comparison with the conventional NPC inverter, the proposed system is 15.25% computationally expensive which yields long execution time and thus less sampling frequency. In this study, two simplified MPC strategies are proposed for the 3L-SNPC inverter system in order to reduce the computational burden: single voltage vector prediction based MPC and selective voltage vector prediction based MPC. Both simplified strategies yield similar performance as the conventional MPC. The required execution times for the simplified MPC strategies are tested on hardware dSPACE 1104 platform. It is found that the single voltage vector prediction based MPC and the selective voltage vector prediction based MPC are computationally efficient by 8.28% and 62.9%, respectively, in comparison with the conventional MPC strategy. However, the average switching frequency and the overall loss in the proposed 3L-SNPC inverter are higher by 83.33% and 46.3%, respectively, than the conventional NPC inverter for a specified load current.
Multi-Mask Based Stabilization of Turbulence Degraded Videos Containing Moving Objects
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladesh, 2019-11) Ray, Bhabesh; Halder, Dr. Kalyan Kumar
Stabilizing videos and detecting moving objects are important tasks in many computer vision applications through it becomes challenging because of the presence of atmospheric turbulence that causes random pixel shifting and blurring of the videos. Because of random change in magnitude and direction of hot air and winds present in the atmosphere, the refractive index of the medium changes nonuniformly. This cause deformation of pixels and makes the tracking system confused while detecting moving objects. This effect becomes more severe with the increase of imaging distance. This thesis proposes an improved method for correcting geometrical distortions of videos degraded by atmospheric turbulence while keeping moving objects unaltered. In this method, by taking the median of input frames, the background frame is estimated and three different techniques are used to generate three different masks. Then, by combining all three masks, a more accurate mask is generated. This refined mask is employed to properly detect the moving objects, and finally combining with the background a stabilized video output is obtained. Performance of this method is tested by applying it on different real-world datasets. A comparison with an existing method shows that the proposed method gives better moving objects detection and improved stabilization of distorted videos.
Analysis of Heavy Metal Concentration in Soils of a Waste Disposal Site in Khulna using Artificial Intelligence Techniques
(Khulna University of Engineering & Technology (KUET), Khulna, Bangladeshtt, 2019-09) Sarkar, Shyamol Kumar; Islam, Prof. Dr. Md. Rafizul
The collection of soil samples is labored and time consuming as well as the determination of heavy metal concentrations in laboratory was expensive. To these attempts, artificial intelligence techniques (AI) such as adaptive neuro-fuzzy inference system (ANFIS), support vector machine (SVM) and artificial neural networks (ANN) were implemented for the analysis of heavy metal concentrations in soils of a selected waste disposal site at old Rajbandh, Khulna. The aim of this study was to fix the functions, algorithms, optimization methods for AI techniques based on their best performance and then select a best technique for the analysis of heavy metal concentrations in soils. In this study, soil samples were collected from eighty-five locations at a depth 0-30 cm from the existing ground surface from the selected disposal site. In the laboratory, the concentrations of heavy metals of Pb, Cu, Ni, Zn, Co, Cd, As, Sc, Hg, Mn, Cr, Ti, Sb, Sr, V and Ba in soils were measured. Result reveals the model with SCP, gaussmf, linear and hybrid was the best-fitted model of ANFIS for the prediction of heavy metal concentrations in soils. In addition, in SVM analysis, the model SVM-RBF with 15 folds was selected for the prediction of heavy metal concentrations in soils. In ANN, the model LT (Levenberg-marquardt and Tansig functions) with neuron structure 2-10-1 was selected. The accuracy of the predicted results were checked based on the acceptable limits of prediction parameters like R value, RMSE, MAPE, GRI and percentage recovery. Among all heavy metals analysis in ANFIS, the maximum R-value 0.999 was found with the minimum RMSE 0.12 for Sc indicating the best correlation in prediction of Sc in soils. The others value of prediction parameters (MAPE= 36.00, GRI=1.50, percentage recovery=123.43%) for Sc were found within the acceptable limits. In addition, in SVM analysis, maximum R-value 0.73 with RMSE 2.03 was found for Cu; while, maximum R-value 0.88 with the minimum RMSE 1.01 for As was found in ANN. The results demonstrated that ANFIS model was a reliable technique than that of other counterparts of SVM and ANN to analyse the heavy metal concentrations in soils with the acceptable degree of robustness and accuracy. Therefore, the performance of AI techniques may be expressed by the sequence of ANFIS > SVM > ANN. Here it can be noted that one can easily be computed the concentration of a particular heavy metal in soils by inserting GPS values (latitude and longitude) only in the developed rule viewer of ANFIS. Therefore, this newly developed model will further be helpful for other researchers in this line to analysis heavy metal concentration in soils of selected waste disposal sites.

M.Sc. Engg.

Browse

Filters

Settings

Sort By

Results per page

Search Results