The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 511570897340 Drone Detection Using Swerling-I Model with L-Band/X-BandRadar in Free Space and Raining Scenario Salman Liaquat, Nor Muzlifah Mahyuddin, Ijaz Haider Naqvi Unmanned Aerial Vehicles, also known as drones, are being utilized increasingly innumerousfields due to their multiple applications. To ensure safe operations, these drones must bedetectedsuccessfully in free space and raining scenarios. However, the current radars that identifylargertargets, such as aeroplanes, may not be helpful as these drones are relativelysmall. Tosuccessfully identify a drone, the radar designer must carefully design the systemusingtheattributes of the target to be picked up successfully by the radar. The characteristics of anX-bandradar would differ significantly from those of an L-band radar. We examine the Swerlingmodelsand their application to drones to simulate the drones' radar cross-section fluctuationusingtheSwerling-I model. The radar range equation uses the signal-to-noise ratio calculatedfromtheReceiver Operating Characteristic curve to detect a drone using L-band and X-band radarsinfreespace and rain scenarios. The analysis reveals that X-band radar experiences greater attenuationin rainy conditions than L-band radar, even though it possesses superior resolutioncapabilitiesatthe same transmitted power. Nevertheless, the selection between these two radar typeshingesonthe particular detection scenario and prevailing environmental conditions. VIS: Vision, Image And Signal Processing1570906422 Detecting Sleep Disorders from NREM Using DeepSDBPLMHaifa Almutairi Sleep disorders have negative effects on human health. Sleep Disorder Breathing(SDB)andPeriodic Leg Movement (PLM) are common sleep disorders that happen duringsleep. Earlydetection of SDB and PLM from Non-Rapid Eye Movement (NREM) can protect patientsfromhypertension and cardiovascular diseases. In this study, we propose a novel deeplearningarchitecture DeepSDBPLM for classifying Normal, SDB and PLMfromNREMusingElectroencephalogram (EEG) and Electromyogram (EMG) signals. Our proposed model istestedinthree different classification problems using ISRUC-Sleep database. The results showthatourproposed model achieves the best result of F1 score as compared to the state-of-the-arttechniques.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 521570907030 Application of Fuzzy Logic in Stock Markets by UsingTechnical Analysis Indicators Mohamad Tarmizi Abu Seman, Leow Kang and Abdul SattarDinInvesting with proper understanding on the stock is risk taking, while investing basedonrumorisgambling. Increase in retail investors during the pan-demic gives an opportunity of a morevolatilemarket, therefore it leads to a need of using proper tools to screen through the stockmarkettofind worthy assets to invest in. Previous research shows that there are possibilitiesof utilizingartificial intelligence techniques in assessing the market performance such as utilizinggeneticalgorithm with moving average convergence-divergence to generate trading signals or usingdeeprecurrent neural network with closing data to predict the next day stock price. Eventhoughthereare good algorithms out there, it has not been utilized and made available to the publictoaccess.Therefore, this project aims to develop an application to screen through the market usingartificialintelligence techniques such as fuzzy logic to find a company which is worthy toinvestin.Historical price data from 100 companies that made up the Kuala Lumpur CompositeIndex(KLCI)is used to assess the performance of the fuzzy logic application developed. Thetechnicalindicators used for the system is RSI, stochastic and MACD. The trading strategyusingthisapplication is to select stocks which have score lower than 0.5 for a buy signal. The resultsalmostachieve the primary objective of generating 70% correct buy signals for short-termtrading. 1570924084 Assessment of Real-World Fall Detection Solution DevelopedonAccurate Simulated-Falls Zaini Abdul Halim, Abdullah Talha Sözer Sozer, Tarik AdnanAlmohamad One of the urgent and popular research areas is wearable devices-based fall detection(FD). Overthe past 20 years, researchers have conducted many experiments in which falls andactivitiesofdaily living were simulated. Researchers inferred that real-world fall data is in-demandratherthan simulated fall data, but this inference still has lack of comparisons. In thisstudy, anassessment of simulated fall dataset and a real-world fall dataset is proposed. Theassessmentinvestigates the efficacy of simulated data for developing an FD solution. Toobservetheeffectiveness of simulated fall, comparisons were conducted between FD methods developedonsimulated and real-world data. The experiments showed that the method with real-worlddataoffered similar performances to the method with simulated data. In contrast to existingsolutions,the provided comparison revealed that accurate simulated data are beneficial to developareal-world FD solution.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 531570911551 Literature Survey on Edge Detection-Based Methods for BloodVessel Segmentation from Retinal Fundus Images Nazish Tariq, Shadi Mahmoodi Khaniabadi, Soo SiangTeoh, ShirLiWang, Theam Foo Ng, Rostam Affendi Hamzah, ZunainaEmbong, Haidi Ibrahim Retinal vessel segmentation is an essential step in the diagnosis of various retinal diseases. Edgedetection-based methods have shown promising results for retinal vessel segmentationduetotheir ability to identify the boundaries of the vessels. In this paper, we surveyed several edgedetection-based methods for retinal vessel segmentation from three main databases: PubMed,IEEExplore, and Google Scholar. The outcomes from the literature search were filteredbasedoninclusion and exclusion criteria. From the selected literature, information about the edgedetectiontechniques, the image datasets used, and the evaluation measures, are extracted. Fromthisliterature survey, we can see that there are many approaches that have been proposedbyre-searchers to segment the blood vessel edges from the retinal fundus images. Most of themareusing the traditional approaches, such as Sobel operators, and Canny edge detector. Recently,deep learning-based approaches have been proposed for this purpose. Some of thecommonlyused databases for retinal fundus images have also been reported in this review. Severalevaluation measures that have been utilized by researchers have also been identified. 1570908296 Optimizing Feature Selection for Industrial Casting Defect Detection Using QLESCA Optimizer Qusay Shihab Hamad, Sami Abdulla Mohsen Saleh, Shahrel Azmin Suandi, Hussein Samma, Yasameen Shihab HamadFeature selection is critical in fields like data mining and pattern classification, as it eliminatesirrelevant data and enhances the quality of highly dimensional da-tasets. This study explorestheeffectiveness of the Q-learning embedded sine co-sine algorithm (QLESCA) for featureselectioninindustrial casting defect detec-tion using the VGG19 model. QLESCA's performance is comparedtoother op-timization algorithms, with experimental results showing that QLESCAoutper-formstheother algorithms in terms of classification metrics. The best accuracy achieved by QLESCAis97.0359%, with an average fitness value of -0.99124. The proposed method providesapromisingapproach to improve the accuracy and reliability of industrial casting defect detectionsystems,which is essential for product quality and safety. Our findings suggest that usingpowerfuloptimization algorithms like QLESCA is crucial for obtaining the best subsets of informationinfeature selection and achieving optimal performance in classification tasks.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 541570917457 Near Infrared Remote Sensing of Vegetation Encroachment atPower Transmission Right-Of-Way David B.L. Bong The event of electricity outage could cause huge financial losses for the industryandinconvenience to the consumers. Vegetation encroachment at the power transmissionright-of-way is one of the main causes. Transmission line fault could occur when a tree fallsintothevicinity of power transmission line. Conventional inspection method such as groundinspectionisthe simplest approach to counter vegetation encroachment. However, technical personnelisrequired to travel on site to perform the inspection manually. This process is often timeconsumingand prone to human error. Airborne Light Detection and Ranging (LiDAR) systemandsatelliteimagery are remote sensing approaches to inspect the power transmission right-of-way. Theseapproaches could reduce reliance on physical site inspection and remove human error. However,large dataset needs to be processed and specialist equipment is needed for this methodwhichalso increases the overall cost. In this research, a simple yet cost effective methodisusedtodetect vegetation encroachment by using near aerial infrared (NIR) image processingapproach.The process is divided into two parts. First, detect the inconspicuous power transmissionlinebyutilizing Radon Transform (RT) in vertical derivative image and detect the peaks of theRadonTransform. Next, detect the vegetation encroachment in the clearance zone by usinggreennormalized difference vegetation index (GNDVI) algorithm to differentiate betweentreesandglassy plains. Preliminary experiment results show a satisfactory performanceindetectingvegetation encroachment the power transmission right-of-way.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 551570897706 A Study of Emotion Recognition Through Heart Rate VariabilitySignals from Low-Cost Device Hermin Kartika Sari, Sunarno Sunarno, Sunu Wibirama, Kwartarini Yuniarti, Rony Wijaya, Memory Motivanisman, WahyuSukestyastama Putra Emotion is a complex state that encompasses feelings, thoughts, and behaviors inresponsetointernal or external stimuli, and is a fundamental aspect of learning, cognition, memory, perception,problem-solving, and human experience. It has a significant impact on human decision-makingprocesses and affects both physiological and psychological states. Due to its importance,researchers have developed various methods to automate the recognition of emotional expression,which has led to numerous studies on the classification of emotions using different combinationsof bio-parameters, emotion classifiers, and accuracy values. In this study, emotion recognitionwillbe classified using machine learning algorithm, namely support vector machine (SVM), K-NearestNeighbor (KNN), and random forest (RF), to classify emotion recognition based onheart ratevariability (HRV) data obtained from a low-cost device. The measurement data servesasthetraining set, while HRV data from the dreamer dataset is utilized as the test set. Thebestclassification outcome is achieved using the random forest algorithm, which yields anaccuracyof77%. EDA: Electronic Design and Application1570909580 Comparison of Different Data Detection Methods in Ortogonal Frequency Division Multiplexing (OFDM) SystemAeizaal Azman Abdul Wahab, Nur Qamarina MuhammadAdnan, Syed Sahal Nazli Alhady Syed Hassan and Wan OthmanOrthogonal Frequency Division Multiplexing (OFDM) is a multicarrier modulation techniquethathasbeen widely used in current technologies due to its many advantages. However, OFDMsuffersfrom high Peak to Average Power Ratio (PAPR) that can distort OFDM's good performance. Tocombat this problem, Selective Mapping (SLM) was used by many researchers but later discoveredthat SLM requires side information (SI) to be transmitted to the receiver and wastes thedatarate.Blind receiver was proposed so that data can be recovered without transmission of SI. Thispaperstudies two of the most famously used blind detectors which is MaximumLikelihood(ML)andViterbi Algorithm (VA) and compares their performance.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 561570917183 Enhance the AlGaN/GaN HEMTs Device Breakdown VoltagebyImplementing Field Plate: A Simulation Study Naeemul Islam, Mohamed Fauzi Packeer Mohamed, MuhammadFirdaus Akbar, Nor Azlin Ghazali, Hiroshi Kawarada, MohdSyamsul, Alhan Farhanah Abd Rahim, AsrulnizamAbdManaf Recently, the features of AlGaN/GaN high electron mobility transistors (HEMT) knownasbreakdown voltage (BV) have garnered a lot of interest for RF and Power applications. But duetothe electric field and current collapse, the break-down voltage of the GaN HEMT deviceisreduced.Therefore, in this research, the field plate technique has been studied for enhancing theGaNHEMTdevice breakdown voltage by using Silvaco TCAD software. It's observed that the dual fieldplatehas a higher breakdown voltage as compared to the gate field plate and source field plate, whichisaround 1100 V. Subsequently, GaN HEMTs presented a threshold voltage (VTH) of -3.3Vandtransconductance (GM) of 16.3 mS/mm approximately. 1570936141 A Comparative Analysis on Electrical and PhotovoltaicPerformances of MIS Structures on High Resistivity SiliconwithTunneling Insulator Nur Bashirouh Attaullah, Nur Zatil 'Ismah Hashim, ChongKahHui,Nor Muzlifah Mahyuddin, Alhan Farhanah Abd Rahim, Marzaini Rashid, Mundzir Abdullah. MIS structures utilizing tunneling AlN on high resistivity silicon is superior in terms of fabricationsimplicity and improved bias responses in addition to providing promising electrical andphotovoltaic performances. Based on previous simulation work, tunneling behavior isabsent fromthe AlN-based MIS photovoltaic properties, indicating inconsistencies with the experimentalevidence. This work aims to highlight this inconsistency by providing a comparativeanalysisbetween AlN and other insulating materials such as SiO2, Si3N4 and Al2O3 in termsof thedarkcurrent, photocurrent and K ratio parameters. Results show that the absence of tunnelingisstillprominent in AlN, whilst the other insulating materials illustrate excellent electrical andphotovoltaic properties evident by the high K ratio values ranging between 103 - 105. Thiscouldbedue to the misrepresentation of AlN in the simulation tool, which requires further parametricadjustments.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 571570917315 Design of Low-Power and Area-Efficient Square Root CarrySelect Adder Using Binary to Excess-1 Converter (BEC) Nor Azlin Ghazali, Mohamed Fauzi Packeer MohamedandMuhammad Firdaus Akbar The Carry Select Adder (CSLA) is commonly used in VLSI design applications like data-processingprocessors, ALUs, and microprocessors to perform fast arithmetic operations. Comparedtoprimitive designs like Ripple Carry Adder and Carry Look Ahead Adder, the regular CSLAoffersoptimized results in terms of area. However, it is still possible to reduce the areaandpowerconsumption of CSLA by implementing a simpler and more efficient gate-level modification. Inthiswork, all the CSLA structures were designed using Verilog HDL while pre-layout simulationandsynthesis were done using Quartus Prime, ModelSim and Synopsys EDA tools. Thefinal resultsanalysis obtained have proven that the BEC-based SQRT CSLA is better than regular squarerootCSLA (SQRT CSLA) as it has reduced total cell area by 19.54% (16-bit) and 19.44%(32-bit) aswellasreduced total dynamic power by 8.52% (16-bit) and 8.75% (32-bit). Ultimately, the modifiedSQRTCSLA structure using BEC method showed significant lower dynamic power consumptionandsmaller cell area than the regular SQRT CSLA. 1570924113 Acoustic Beamforming Using Machine Learning Te Meng Ting, Nur Syazreen Ahmad This paper shows how two microphones in an endfire array configuration was usedtoperformbeamforming. The setup uses two condenser microphones and a sound card to allowmultiplesources to be input to the computer at the same time. A cross-correlation calculationwasusedtodetermine the time shift between the two mics. Using the Delay and Sumalgorithm, thetimeshiftcan be corrected, and the mic signals can be added to a superposition. 1570922646 Simulation of Bottom-Gate Top-Contact Pentacene BasedOrganicThin-Film Transistor Using MATLAB Nor Azlin Ghazali Organic transistor plays an important role in electronic applications as it providesadditionalbenefit of flexibility and low cost compared to silicon electronic devices. Simulation andanalyticalmodel of such organic devices helps in improving and optimizing the performance of thedevice.There are various parameters that effect the performance of the device. In this work, analyticalsimulation of the organic device is performed based on the device physics to simulatetheoutputcharacteristics using MATLAB for various channel length (L). Here, the impact on channel lengthisobserved on varying the device length and its impact is observed on the drain current. Differentchannel length taken into consideration and the simulation result shows that the draincurrentincreases when the channel length decreased. Thus, simulation of such analytical modelshelpsinextracting the useful information about the performance of the organic transistors.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 581570918394 A Comparative Study of Noise Reduction Techniques for BloodVessels Image Shadi Mahmoodi Khaniabadi, Haidi Ibrahim, Ilyas AhmadHuqqaniand Harsa Amylia Mat Sakim, Soo Siang Teoh The accurate analysis and interpretation of blood vessel images are essential for diagnosingandmonitoring various medical conditions. However, these images often suffer fromthepresenceofnoise, which can hinder proper visualization and lead to erroneous interpretations. Inthispaper,we present a comprehensive comparative study of noise reduction techniques for bloodvesselimages. The study encompasses both traditional and new methods, evaluating their performance,benefits, and challenges. Traditional methods, such as Anisotropic Diffusion FilteringandWaveletTransform, have proven effective in preserving blood vessel structures and retainingfinedetails.However, they require careful parameter selection and may be computationally intensive. Ontheother hand, new techniques, including Contrast Limited Adaptive HistogramEqualization(CLAHE),Non-Local Mean Filter (NLM), and deep learning-based approaches, offer promising advancementsin noise reduction capabilities with reduced computational complexity. The choicebetweentraditional and new methods depends on specific application requirements, noise characteristics,and available computational resources. Our findings highlight the need for further researchinparameter tuning, computational efficiency optimization, and hybrid approaches toenhancethenoise reduction process in blood vessel images. This study contributes to the advancementofmedical imaging by providing valuable insights for researchers and practitioners, enablingimproved diagnostic accuracy and patient care. TECHNICAL SESSION 4 VIS: VISION, IMAGE AND SIGNAL PROCESSING1570907566 YOLOv7-Tiny and YOLOv8n Evaluation for Face DetectionIbrahim Ali Ahmed Abdullah AL-Amuodi, Dzati Athiar Ramli The lightweight face detection models were developed to match the specifications of edgedevices.This makes them suitable for use in real-life applications where the high GPUmight notbeavailable. This study compares the performance of two lightweight object detectionmodels,YOLOv7-tiny and YOLOv8n, for face detection applications. Both models weretrainedonWIDERFACE dataset, and their performance was evaluated using mean average precision(mAP).Regarding model size, YOLOv8n is lighter than YOLOv7-tiny with only 3.01 millionparameters,making it more efficient for deployment on re-source-constrained devices. Regardingaccuracy,the results showed that YOLOv7-tiny was higher in mAP50 and YOLOv8n Was higher inmAP50-95.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 591570925412 Wood Defect Inspection on Dead Knots and Pinholes UsingYOLOv5x Algorithm Pei Yi Liew, Muhammad Firdaus Akbar, Bakhtiar Affendi Rosdi, Muhamad Faris Che Aminudin, Mohd 'Akashah FauthanAccurate detection of wood defects is crucial for ensuring the quality and reliability of woodpiecesin various industries such as construction and furniture production. Some challengingdefectssuch as dead knots and pinholes vary in size and shape, complex textures, and thepresenceofwood grains makes the inspection process more complex. Thus, this paper evaluatestheperformance of the YOLOv5x algorithm in detecting and localizing wood defects, especiallydeadknots, and pinholes. Using an augmented custom dataset and trained with transfer learningfromapre-trained model with a COCO dataset, the algorithm achieves a precision score of 96.5%, arecallscore of 91.8%, and a [email protected] score of 95.5%, it indicates highly accurate defect detectionandlocalization. Visual examples demonstrated the algorithm's capabilities as well as instancesofincorrect detections and failed detections. The findings of this study can contribute tothefieldofwood inspection systems and highlight areas for further improvement in defect detectionalgorithms. 1570924384 Deep Learning Based Distance Estimation Method UsingSSDandDeep ANN for Autonomous Braking/Steering Siti Nur Atiqah Halimi, Mohd Azizi Abdul Rahman, HattaAriff, YapHong Yeu, Nor Aziyatul Izni Mohd Rosli, Mohd AzmanAbas, SyedZaini Putra Syed Yusoff The Automatic Emergency Braking (AEB) system is a mechanism that enables driverstoleveragethe capabilities of their vehicles by warning them of potential collisions and assistingtheminaverting them. Autonomous Emergency Steering (AES) is one of the active safety systemsthatcanassist with evasive steering. It will make it simpler for the driver to avoid an accident that couldhave been prevented. Concerns include the distance necessary to prevent a collisionwhenturningor reversing and the required space when braking and turning. Given such inquiries, developingasystem to estimate the distance between the vehicles is necessary. Consequently, thisstudysuggested utilizing deep learning for AEB and AES to estimate the distance between vehiclesusinga monocular vision sensor. In addition, the object distance estimation method is employedasadistance estimation method. Experiments are conducted to determine the precisionoftheproposed method for estimating the distance between the target vehicle and the camerausingLiDAR distances. The result indicates that the proposed method for estimating distancehasanaccuracy of 92% compared to LiDAR distance. As a result, the findings of this researchhavethepotential to contribute to the methodological foundation for further understanding drivers' behavior,with the ultimate objective of lowering the number of accidents involving rear-end crashes.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 601570908304 Improving the Accuracy of Gender Classification BasedonSkinTone Using Convolutional Neural Network - Transfer Learning(CNN-TL) Muhammad Firdaus Mustapha, Nur Maisarah Mohamad, Siti Haslini Ab Hamid Gender classification is one of the key features in soft biometrics besides age, ethnicity, facialexpression, etc. Gender classification based on skin tone has its own importance that canfurtherimprove the performance of facial recognition systems. Most CNN models require a largeamountof training data to improve classification accuracy and increase processing time. Fortunately, theMobileNetV2 model overcomes this problem by running faster than the other models. However, themodel's accuracy suffers when the gender classification results based on skin tonereach50%accuracy, indicating that the model suffers from an "overfitting" problem. To address thisissue, theproposed research constructed a novel face images dataset containing 6250 faceimagesthatchosen from original FaceARG dataset and divided equally into Bright (3125) and Dark(3125)skintone. Each skin tone (Bright and Dark) is equally divided into two genders, 1563 (male) and1563(female). The new FaceARG dataset is then used to run two types of experiments (Bright andDark)on the MobileNetV2 model. The Fine-Tuning method from Transfer Learning is thenappliedtotheMobileNetV2 model, along with method gains from previous studies. The Dark experiment achievedthe highest accuracy on training dataset, which is 97.4%, compared to 50% on the model withoutfine-tuning, and the Bright experiment achieved the highest accuracy on test dataset, whichis89.8%, compared to 50% on the model without fine-tuning. The findings of this study will determinethe ability of the proposed model to accurately classify gender based on skin tone. 1570917574 Evaluation of Three Variants of LBP for Finger CreasesClassification Ahmad Nazri Ali, Imran Riaz, Nur Azma Afiqah SalihinBiometric technology improves security and authentication, especially in sensitive systemslikeattendance systems. The common traits for biometrics are typically fromthe iris, fingerprints, face,etc. Another trait that possibly comes from the finger creases and the research workevaluatingthe finger crease's capability for biometric classification is proposed in this paper. Variouslocalbinary patterns (LBP) are employed to extract the features, and the classification performanceisevaluated using Support Vector Machines (SVM) on two different kernels. Fromthe evaluation, anaccuracy of up to 94% with a percentage of FAR and FRR less than 3 is observedforall theproposed LBP methods.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 611570917772 R-Peaks and Wavelet-Based Feature Extraction on K-NearestNeighbor for ECG Arrhythmia Classification Ku Nurul Fazira Ku Azir, Adam Mohd Khairuddin, MohdRashidi Che Beson The aim of this research is to classify 17 types of arrhythmias by applying the algorithmdevelopedfrom combining the morphological and the wavelet-based statistical features. Theproposedarrhythmia classification algorithm consists of four stages: pre-processing, detectionof R-peaks,feature extraction, and classification. Seven morphological features (MF) that were retrievedfromthe R-peak locations. Following this, another nine wavelet-based statistical features(SF) weregathered by decomposing wavelets in level 4 from the Daubechies 1 wavelet (Db1). These16features are then applied to the k-nearest neighbor (k-NN) algorithm. The accuracy (ACC) ofthesuggested classification algorithm was assessed by using the MIT-BIH arrhythmiabenchmarkdatabase (MIT-BIHADB). The experimental results of this work attained an average accuracy(ACC)of 99.00%. 1570921465 A Finger Knuckle Print Classification SystemUsing SVMforDifferent LBP Variants Ahmad Nazri Ali, Imran Riaz, Ilyas Ahmad Huqqani Finger knuckle print is one of the most important biometric traits and plays a vital roleinasecureidentification system. In this paper, performance evaluation of local binary pattern(LBP) anditsvariants center symmetric local binary pattern (CS-LBP) and median local binary pattern(MLBP)are investigated. After feature extraction, a support vector machine (SVM) with the linear kernelisused for the performance evaluation of two different datasets named the Poly-UFKPdatasetandthe USM-FKP dataset. The experimental results show that CS-LBP performs better for theUSM-FKP dataset with an accuracy of 86.2% which demonstrates the potential of the FKPclassificationsystem.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 621570919241 Survey on Blood Vessels Contrast Enhancement AlgorithmsforDigital Image Shadi Mahmoodi Khaniabadi, Harsa Amylia Mat Sakim, Haidi Ibrahim, lyas Ahmad Huqqani, Farzad Mahmoodi Khaniabadi, SooSiang Teoh This paper surveys blood vessel contrast enhancement algorithms in digital images, aimingtooptimize imaging techniques for accurate analysis and interpretation of vascular structures.Various contrast enhancement techniques, including global and local approaches, areemployedtoimprove the visibility and differentiation of blood vessels from the surrounding background. Theinvestigation reveals that both global and local enhancement techniques play vital rolesinenhancing blood vessel contrast. Global enhancement methods, such as spatial andfrequencydomain approaches, focus on enhancing overall contrast and visibility throughout the entireimage.Yet, local enhancement techniques selectively enhance contrast and visibility in specificregionsofinterest, while preserving overall image quality. By combining global and local enhancementapproaches, researchers can achieve comprehensive and targeted enhancement of bloodvesselvisibility and analysis. The findings emphasize the significance of utilizing suitable enhancementtechniques to optimize blood vessel contrast in digital images and advance the fieldof medicalimaging. This research contributes valuable insights for the development of optimizedimagingtechniques and algorithms for accurate blood vessel analysis and diagnosis. 1570920593 Face Image Authentication Scheme Based on Cohen-Daubechies-Feauveau Wavelets Muntadher H. Al-Hadaa, Rasha Thabit, Khamis A. Zidan, BeeEeKhoo The recent interest of face image manipulation detection has been di-rected towards providingtheability of detecting various types of manipulations. In the best scenario, the availablemethodscandetect the manipulations and lo-calize the manipulated face region. The ability of recoveringtheface region af-ter manipulation localization will be very useful in practical applications, how-ever,this has not been highlighted in the previous researches. In this paper, a newfaceimageauthentication (FIA) scheme is presented based on image wa-termarking and Cohen-Daubechies-Feauveau (CDF) wavelets. In the proposed scheme, the CDF is used to generate the recoverybitsfrom the face region in order to be used for recovering the face region when manipulationsexist.Sev-eral experiments have been conducted to evaluate the performance of the pro-posedschemewhich proved its efficiency in generating high quality water-marked images, detectingvarioustypes of manipulations, localizing the manip-ulated blocks in the face region, and recoveringtheface region with good visual quality. The comparison with the state-of-the-art detectionschemesproved the superiority of the proposed scheme.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 631570918314 Ground Truth from Multiple Manually Marked Images toEvaluateBlood Vessel Segmentation Nazish Tariq, Michael Chi Seng Tang, Haidi Ibrahim, SooSiangTeoh, Zunaina Embong, Aini Ismafairus Abd Hamid, RafidahZainon Blood vessel segmentation from digital images is one of the valuable processesfor medicaldiagnosis. Many researchers have proposed blood vessel segmentation algorithms, whichcansegment the blood vessels automatically or with minimum human interventions. Oneofthepopular blood vessel segmentation branches is edge-based segmentation. In this approach, onlythe edges are detected by the algorithm. While developing edge segmentationalgorithms,researchers must evaluate their proposed methods' performance. If full-reference-basedqualitymeasures are utilized, the ground truth, which shows the targeted segmentation output, isneeded. This ground truth is commonly generated manually, where human experts identify anddrawtheedges. However, the manually segmented edges may differ depending on the expertsduetoseveral factors, including individual preference. The work in this paper aims to givesomeinsightinto how to combine these images. We suggest that the edges be classified as useful edges, weakedges, and unintentional edges.
The 12th International Conference on Robotics, Vision, Signal Processing & Power Applications28 & 29 August 2023 64The organizing committee acknowledges the efforts of all thosewhohave contributed their valuable time and efforts as reviewersinensuring high quality technical papers for ROVISP2023. Inadditionto this, to the International Advisory Committee, thankyouforthecontinuous support to this conference. Deepest appreciation is due to all staff of the School of Electricaland Electronic Engineering, Universiti Sains Malaysia(USM). Wewould like to express our gratitude to Springer for their technicalsupports, as well as STMicroelectronics, our gold sponsor, CADFEM,our silver sponsor, Penang Convention and Exhibition Bureau (PCEB) for their financial supports. Subsequently, the organizing committee would like toexpressourutmost thanks to the Deputy of Secretary General of MOSTI, DatukTs. Dr. Mohd Nor Azman bin Hassan for officiatingtheopeningceremony of this event. Finally, we would like to extend our heartfelt gratitudetoall theparticipants, attendees and exhibitors during ROVISP2023. ACKNOWLEDGEMENTS