NUST Institutions Library Catalogue Search for 'an:123222'

By Zia ur Rehman, Muhammad . . 153p. , Advancement in the myoelectric interfaces have increased the use of myoelectric controlled robotic arms for partial-hand amputees as compared to body-powered arms. Current clinical approaches based on conventional (on/off and direct) control are limited to few degree of freedom (DoF) movements which are being better addressed with pattern recognition (PR) based control schemes. Performance of any PR based scheme heavily relies on optimal features set. Although, such schemes have shown to be very effective in short-term laboratory recordings, but they are limited by unsatisfactory robustness to non-stationarities (e.g. changes in electrode positions and skin-electrode interface). Moreover, electromyographic (EMG) signals are stochastic in nature and recent studies have shown that their classification accuracies vary significantly over time. Hence, the key challenge is not the laboratory shortterm conditions but the daily use. Thus, this work makes use of the longitudinal approaches with deep learning in comparison to classical machine learning techniques to myoelectric control and explores the real potential of both surface and intramuscular EMG in classifying different hand movements recorded over multiple days. To the best of our knowledge, for the first time, it also explores the feasibility of using raw (bipolar) EMG as input to deep networks. Task are completed with two different studies that were performed with different datasets. In the first study, surface and intramuscular EMG data of eleven wrist movements were recorded concurrently over six channels (each) from ten able-bodied and six amputee subjects for consecutive seven days. Performance of stacked sparse autoencoders (SSAE), an emerging deep learning technique, was evaluated in comparison with state of art LDA using offline xii classification error as performance matric. Further, performance of surface and intramuscular EMG was also compared with respect to time. Results of different analyses showed that SSAE outperformed LDA. Although there was no significant difference found between surface and intramuscular EMG in within day analysis but surface EMG significantly outperformed intramuscular EMG in long-term assessment. In the second study, surface EMG data of seven able-bodied were recorded over eight channels using Myo armband (wearable EMG sensors). The protocol was set such that each subject performed seven movements with ten repetitions per session. Data was recorded for consecutive fifteen days with two sessions per day. Performance of convolutional neural network (CNN with raw EMG), SSAE (both with raw data and features) and LDA were evaluated offline using classification error as performance matric. Results of both the short and long-term analyses showed that CNN and SSAE-f outperformed the others while there was no difference found between the two. Overall, this dissertation concludes that deep learning techniques are promising approaches in improving myoelectric control schemes. SSAE generalizes well with hand-crafted features but fails to generalize with raw data. CNN based approach is more promising as it achieved optimal performance without the need to select featur 30cm..

Place Hold on Deep Learning for Improved Myoelectric Control /

A Robust Scheme of Vertebrae Segmentation for Medical Diagnosis /

. 114p. , Automated vertebrae analysis from medical images plays an important role in computer aided diagnosis (CAD). It provides an initial and early identification of various vertebral abnormalities to doctors and radiologists. Vertebrae segmentation and classification are important but difficult tasks in medical imaging due to low contrasts in image, noise and high topological shape variations in radiological scans. It becomes even more challenging when dealing with various deformities and pathologies present in the vertebral scans like osteoporotic vertebral fractures. In this work, we want to address the challenging problem of vertebral image analysis for vertebra segmentation and classification. In the past, various traditional imagery techniques were employed to address these problems. Recently, deep learning techniques have been introduced in biomedical image processing for segmentation and characterization of several abnormalities. These techniques are becoming popular in solving various medical image analysis problems due to their robustness and accuracy. In this research, we present a solution of vertebrae segmentation and classification problem with the help of deep learning approach. We present a novel combination of traditional region based level-set with deep learning framework in order to extract the shape of vertebral bones accurately; which would be able to handle the deformities in the vertebral bones precisely and efficiently. After vertebrae segmentation, we further extend the work to abnormal vertebrae classification with the help of deep learning approach. This novel framework would be able to help the medical doctors and radiologists with better visualization of vertebral bones and providing the initial automated classification of vertebrae to be normal or abnormal. The proposed method of vertebrae segmentation was successfully tested on different datasets with various fields of views. The first dataset comprises of 173 CT scans of thoracolumbar (thoracic and lumbar) vertebrae in sagittal view, collected from a local hospital. The second dataset comprises 73 CT scans of cervical vertebrae in sagittal view, also collected from a local hospital. The third dataset comprises 20 CT scans of thoracolumbar (thoracic and lumbar) vertebrae in sagittal view collected from spine segmentation challenge CSI 2014. The forth dataset comprises 25 CT scans of lumbar vertebrae in sagittal view collected from spine segmentation challenge CSI 2016. Lastly, we have utilized the same locally collected set of 173 iii CT scans of thoracolumbar (thoracic and lumbar) vertebrae and extracted in axial view to perform the segmentation task. For classification purpose, we have utilized the locally collected set of 173 CT scans of thoracolumbar (thoracic and lumbar) vertebrae as these include osteoporotic vertebral fractures in it. The details of these datasets have been presented in respective sections. We have achieved promising results on our proposed techniques. The evaluation of the segmentation performance on the datasets with various machines and field of views helped us to ensure the robustness of our proposed method. On validation sets of these datasets, we have achieved an average dice score of around 95% for vertebrae segmentation; and accuracy of above 80% for the vertebrae classification. The detailed results have been presented in the results section. These results reveal that our proposed techniques are competitive over the other state of the arts in terms of accuracy, efficiency, flexibility and time 30cm.

Place Hold on A Robust Scheme of Vertebrae Segmentation for Medical Diagnosis /

Multi-Robot Interactive Therapy for Children with Autism Spectrum Disorder /

By Ali ,Sara . . 139p. , Robot-Assisted Therapies (RAT) is an emerging field and has shown promising results. Recently, one of the prominent applications of robots is assistive therapy is for Autism Spectrum Disorder (ASD). Autism Spectrum Disorder is a set of neurodevelopment disorder affecting 7.5 million people around the world. Children with ASD lack social and communication skills which affects their ability in schools as well as in community. Recently, humanoid robots are used for the treatment of children with autism spectrum disorder to improve the development of communicational, behavioral, motor movements, joint attention, and physical behaviors. These interactive interventions that use robots for children with ASD, is one of the favorable tools for improving the behavior of children. In particular, the area of robotics is helping a lot in the treatment of ASD as the robot acts as a mediator as well as measures the response of an autistic child. However, the research aiming that the treatment of children with autism is limited, these therapies introduced by robots are successful in establishing basic communication skills. This research has proposed a novel mathematical model for an adaptive therapy of children with Autism Spectrum Disorder called Multi-robot-mediated Intervention System (MRIS). Three different therapies related to improvement in joint attention and imitation skills, effective human-human interaction and comparison of effective stimulus are introduced under this mathematical model. This research aims to introduce multi-sensory data that provides the quantitative support for improvement in social skills of children with autism, replacing the current techniques of measuring the improvement from physically observing the ASD child and with video analysis. Besides ensuring the accuracy in results, this method also introduces consistency as robots are immune to fatigue, unlike humans. The effectiveness of the model has been validated using cognitive brain state of the children with Electroencephalogram (EEG) neuroheadsets. Moreover, the effectiveness of the results has been validated using statistical analysis and the Childhood Autism Rating Scale (CARS). 30cm.

Place Hold on Multi-Robot Interactive Therapy for Children with Autism Spectrum Disorder /

Multi-Gesture Decoding using Hybrid EMG-IMU for Rehabilitation Applications /

By Shahzad, Waseem . . 161p. , Prosthetic rehabilitation of the upper limb through powered prosthetic devices is an active research area for the past several decades. Multi-degree of freedom prosthetic hands controlled by surface electromyography signals (sEMG) of the forearm muscles have been developed to restore the lost limb functionality. A major objective of the current research is the development of intuitive prosthetic controllers for modern dexterous prosthetic hands. Current state of the art prosthetic controllers utilize pattern recognition (PR) of multi-channel sEMG signals for decoding the intended motion class. Despite decades of academic research, a clinically viable PR based prosthetic controller is yet to be realized. The PR based classifiers have resulted in more than 90% classification accuracy when evaluated under controlled laboratory conditions. However, the excellent laboratory performance of these classifiers is yet to result in their clinical acceptability and commercial availability. As a result, the currently available prosthetic devices are still based on non-intuitive binary or sequential digital control. There are several reasons for this academia-industry disparity. The unintended variations of sEMG signal characteristic due to several confounding factors, including arm position, adversely affects the performance of pre-trained PR based prosthetic controllers. Researchers have proposed fusion of auxiliary sensory information for classifier robustness against forearm positional variations. Sensor fusion techniques including accelerometer-mechanomyography (ACC-MMG), force-myography (FMG) and magnetic markers (MMs), have been reported with significant performance improvements. This study focused on the impact of arm position variations on the performance of PR-based forearm motion class decoders. A wearable data acquisition system was designed to acquire multichannel sEMG and measure arm position using inertial measurement units. The performance of support vector machine (SVM) and linear discriminant analysis (LDA) classifiers was evaluated for training at static positions and for dynamic arm movements to characterize the adverse effects of arm position variation. Sensor fusion of sEMG and arm position data was evaluated to mitigate the arm position effect. A comparison of static multi-position training and dynamic arm movement training was carried out to suggest more pragmatic strategies for classifier training. The sensitivity xi of classifiers to motion class taxonomy was also evaluated. Sixteen motion classes categorized in distant-taxonomy and close-taxonomy groups were classified using LDA, Linear SVM (LSVM), Non-Linear SVM (NLSVM) and Multi-layer perception (MLP) classifiers. The results of the study have shown a significant dependence of LDA and LSVM classifiers and an insignificant dependence of the NLSVM and MLP classifiers on motion class taxonomy. The results suggest a more pragmatic selection of motion classes for realistic classifier performance evaluation 30cm.

Place Hold on Multi-Gesture Decoding using Hybrid EMG-IMU for Rehabilitation Applications /

Brain-Computer Interface for Mental State Detection of Drivers /

By Arif, Saad . . 169p. , Background: Each year millions of vehicles suffer crashes on the roads globally due to deteriorated mental state of the drivers during driving tasks which result in higher casualties. Drowsy driving is the leading cause of high fatality rate which is instigated due to sleep deprivation, fatigue, and anxiety, etc. Vehicular, and driver’s behavioral data-dependent systems detect the drowsiness after its onset when an accident is more likely, and they are also subject to false identification. The drowsy mental state of drivers must be detected earlier for in-time warning to avoid fatal losses, and also the system must be less intrusive, and adaptable for normal driving tasks. Detection systems using physiological signals from human organs are comparatively underexplored in which bodily states of the subject can be identified earlier and with more reliability. It is postulated that all the bodily states are primarily originated from the human brain which could be a promising region to detect drowsiness at an earlier stage. Among many physiological signals, electroencephalography (EEG), and functional near-infrared spectroscopy (fNIRS) are chosen for this study because they are proved to be more portable, easy to use, non-invasive, and reliable brain modalities for human mental state detection. Aim and Objectives: This study aimed to design a passive brain-computer interface (pBCI) scheme for the earlier detection of driver drowsiness with minimum intrusion into the driving task. Design objectives for the pBCI scheme were to select the channel of interest (COI), online detection with a shorter time window, minimum recalibration and setup time, and development of a widely applicable standard as an inter-subject transfer framework (iSTF) to cater to the inter-subject variability. All these objectives lead to such a pBCI system that is readily available for any subject, anytime, easy to use with minimal design, and yet detecting the drowsiness correctly at an earlier stage to avoid life losses. Methodology: Multichannel EEG and fNIRS brain signals from anterior, posterior, and lateral brain regions of sleep-deprived drowsy subjects were acquired during the simulated driving task for post hoc analysis. Initial pBCIschemes used labeled EEG data acquired from prefrontal (PFC), frontal, and occipital cortices for extracting the eight spectral, and eight temporal xvi features of EEG signals. Seven supervised machine learning classifiers were used to do the cross-validated binary classification of drowsy, and alert brain states. Initial design only achieved a few objectives and generated the need to use different modalities to meet all the requirements. The final pBCI scheme used labeled fNIRS data acquired from PFC and dorsolateral prefrontal cortices (DLPFC) for extracting the six cerebral oxygen regulation (CORE) and three hemodynamic signal features. CORE states of wakefulness and non-rapid eye movement (NREM) sleep stages were used to design a novel standard framework for wide applicability, and sleep stage classification using the vector phase analysis (VPA) approach. VPA was used for classifying microsleep/lapse, and drowsiness detection was done using the proposed brain hemodynamic patterns. Results: 𝛿, 𝜃, 𝛼, 𝛽 band powers as EEG spectral features achieved 82% accuracy in 10 s detection window. Signal skewness, variance, mean, peak as EEG temporal features achieved 87.2% accuracy in 1 s detection window. Ensemble classifier declared F8 as COI for earlier drowsiness detection using both the EEG pBCI schemes. Only the objectives of the short detection window and COI were achieved with the initial designs. In the final design with fNIRS, the novel VPA features: CORE vector gradients, achieved 94.1% accuracy in 5 s detection window for NREM sleep stage classification using ensemble classifier with the least computation time of 44 ms. Precise spatial localization of fNIRS declared AF8 position in right DLPFC as COI. The novel sleep stages-based threshold criteria along with VPA were crossvalidated as a standard iSTF for online microsleep detection with the least recalibration and setup time. Feature selection and achieved results were validated with various statistical significance tests. All the design objectives were attained with the fNIRS-based pBCI scheme. Conclusion: The aim to detect the driver's drowsiness earlier with minimum intrusion into the driving task, is accomplished. The presented fNIRS-based adaptive pBCI scheme is readily available for any subject, anytime, easy to wear with minimal ergonomic design, capable of real-time, correct, and early detection of the driver drowsiness to lessen the life losses in vehicular driving scenarios. The recommended research directions will surely justify, improve, and broaden the application horizons of the presented design 30cm..

Place Hold on Brain-Computer Interface for Mental State Detection of Drivers /

DETECTION OF THYROID DISEASES USING MACHINE LEARNING TECHNIQUES /

By AKHTAR,TEHSEEN . . 137p. , Background: The unusual growth of the glandular tissue on the boundary of the Thyroid gland is an indication of Thyroid disease. Thyroid disease is characterised by an unusually high or low number of hormones produced by the thyroid gland, the two most prevalent kinds are hypothyroidism (underactive thyroid gland) and hyperthyroidism (overactive thyroid gland). The main aim of this project was to introduce the concept of an efficient multi-stage ensemble i.e., the voting ensemble of the homogeneous ensemble which could be used with a variety of feature-selection algorithms for improving the diagnosis of thyroid diseases. The dataset utilised in this study was built from real-time thyroid data obtained from the teaching hospital in DG Khan at District Head Quarter (DHQ), Pakistan. Following the appropriate preprocessing processes, three kinds of attribute-selection strategies were used: The first approach used was Select from Model (SFM), the second technique was the Select K-Best (SKB), and the final methodology was the Recursive Feature Elimination (RFE). Select From Model (SFM) is a form of attribute-selection strategy that uses a model to select attributes. As potential feature estimators, the Decision Tree (DT), Logistic Regression (LR), Gradient Boosting (GB) and Random Forest denoted as the (RF) classifiers were employed in conjunction with each other. The homogeneous ensemble activated the bagging, boosting-based learners, who were then classified by the Voting ensemble, which employed both soft and hard voting to categorise the data. Other performance assessment criteria such as hamming loss, accuracy, mean square error, sensitivity and others have been implemented. The results of the experiments reveal that when the suggested approach for better thyroid sickness detection is applied in its most practicable form, it is most successful. On the dataset 1, all of the algorithms tested obtained 100 % accuracy with subset of the total no of feature in each case, however on the dataset 2, more than 98 percent accuracy was reached in every case. On the basis of accuracy and computing cost, the results given here exceeded equivalent benchmark models in their respective fields of study. 30cm.

Place Hold on DETECTION OF THYROID DISEASES USING MACHINE LEARNING TECHNIQUES /

DETECTION, ESTIMATION AND FORECAST FOR NONLINEAR SYSTEMS /

By Zaheer Ud Din, Asim . . 142p. , This thesis presents and implements detection, estimation, and forecast algorithms in context of smart infrastructure system for smart grid. A novel application of radio frequency wireless mesh network and general packet radio service technologies in a telemetry solution has been proposed. The telemetry solution measures power flow in the energy network of an electricity distribution company. The solution utilizes some selected circuits of grid stations, and calculates total power consumed, total power imported and total power exported by the distribution company. The selection of circuits for sensors installation is the key for reducing solution cost as compare to the case when sensors are installed on all the power output points. The framework involves installation of specially developed energy sensors (smart energy meters) and data concentrator units at the selected grid stations. The approach has been tested on two electricity distribution companies of Pakistan: Islamabad Electric Supply Company and Peshawar Electric Supply Company. Also in this work, result of over-load detection based on generalized likelihood ratio test for an industrial feeder of Islamabad Electric Supply Company is included. Detection probability of 0.96 with a false alarm probability of 0.04 has been achieved for 30 minutes data interval. Further, 4 years power data obtained from above mentioned system is utilized in a multivariate multi-step-ahead short-term forecasting formulation. The formulation operates on multiple inputs from multiple variables, and provides multi-step-ahead forecasts by generating multiple outputs for multiple variables. The presented framework is effective for large forecasting horizons since it forecasts for temporally dependent sub- xiii intervals called runs from large horizon. Thus the framework forecasts are less biased and suffer low variance, as compared with direct method and iterated method estimators respectively. Feedward neural network and Long-short-term-memory network models have been evaluated in presented framework. The proposed framework has demonstrated forecasts of power import and power export with a horizon value of 48 for Peshawar Electric Supply Company (PESCO), Pakistan. The averaged mean absolute percentage error of two forecasted time series is 12.76 %, whereas, 24 hours ahead power consumption of PESCO total consumers has been forecasted with mean absolute percentage error of 8.6%. Furthermore, exploiting 24 hours ahead power consumption forecasts has resulted in better power dispatch for PESCO grid stations by reducing mean absolute error by 11.52 times between PESCO power allocated and PESCO power consumed. Next the thesis presents an Euler approximate discrete-time Sliding Mode observer (SMO) which simultaneously estimates states and combined effect of unmodeled system dynamics and disturbances. Emulation Design procedure is employed in designing of discrete feedback linearization controller. Computer simulations demonstrate performance of presented output feedback scheme for tracking applications of magnetic levitation and DC motor systems. Results illustrate that reducing sampling period more adversely affects Euler approximate discrete observer performance for faster changing system dynamics than for slower changing dynamics. The proposed scheme also exhibits good performance in presence of disturbances and parameters perturbation. Furthermore, it is demonstrated via simulations that robust tracking control is achived on using estimator (e.g Kalman filter, SMO, SSRLS filter) in sampled-data output feedback configuration, as compared to performing tracking using sampled-data state feedback scheme. Simulation results show that SMO based output feedback tracking is most robust, followed by CKF and EKF based output feedback scheme. UKF based output feedback scheme is robust against external disturbance; but for case of system parameter perturbation, UKF tracking error takes longer time to converge. State-Space Recursive Least Squares (SSRLS) based scheme behaves poorly in presence of external disturbance. This is because SSRLS estimation is based on constant velocity model and not on actual nonlinear system model. xiv Finally, output feedback control scheme for case of unknown system parameters has been presented. The scheme employs dual UKF estimation algorithm and Emulation Design based discrete feedback linearization controller. Implementation results exhibit that presented output feedback control scheme demonstrates better tracking performance and parameter estimation error when parameter estimate is initialized with a value (in dual estimation algorithm) which is closer to actual system parameter value. 30cm..

Place Hold on DETECTION, ESTIMATION AND FORECAST FOR NONLINEAR SYSTEMS /

TRAJECTORY TRACKING FOR AGRICULTURAL DYNAMIC MULTI COPTER AERIAL ROBOT /

By Nashit Arshad , Syed . . 161p. , Unmanned aerial vehicles (UAVs) have become a popular choice for spraying pesticide in agricultural use due to their versatility and maneuverability. Quadcopters carrying suspended water containers are widely used for firefighting services. The efficient transportation of liquids by UAVs is of utmost importance in various autonomous missions, including agriculture field spraying. A lot of research is being carried out on the control of these UAVs subject to the constraints of unwanted forces created by the sloshing liquid. However, the complex dynamics of this system can result in the degradation of flight safety due to the linkage among the UAV maneuver, container swing, and liquid sloshing. Liquid sloshing in a container is a well-known and longstanding challenge within the field of engineering. In this study, the word liquid sloshing refers to the variable wave surface elevation of the fluid in a container. Nevertheless, liquid sloshing can lead to undesirable effects such as instability, unwanted forces, position error, and increased control effort resulting in inefficient power utilization and payload constraints. To ensure the effective implementation of the control system for an agricultural spraying drone, it is essential to estimate the pesticide slosh model. The objective of this study is to ascertain sloshing parameters by employing an innovative technique that leverages a costeffective sensor. The proposed experimental setup employed during this investigation comprises a rectangular beaker positioned on a conveyor belt. A Kalman estimator based ultrasonic sensor, mounted atop the liquid-filled container whose slosh parameters necessitate identification, is employed. System identification techniques were employed to derive the system model. Comparative analysis involving calculation of the Root Mean Square Error (RMSE) were conducted to evaluate accuracy and error. Following numerous tests conducted at various slosh levels, the acquired data was subjected to analysis. The results obtained substantiate the feasibility of our concept in measuring slosh under dynamic conditions. To mitigate the effects of liquid sloshing, an approach based on Lagrangian is utilized that enables the development of dynamic model of UAV and resulting nonlinear coupled dynamics of liquid carrying quadrotor. This developed hybrid model, incorporating both slosh and drone dynamics, is thoroughly examined. It enables the application of different control strategies to attain satisfactory performance and meet energy requirements based xii on actuator control efforts. The study delves into two specific control methods: Linear Quadratic Regulator (LQR) and Proportional-Integral-Derivative (PID), extensively presenting, investigating, validating, and comparing their effectiveness in achieving stability and calculating energy demands for a hovering liquid-carrying quadcopter. The utilization of LQR and PID controllers offers notable enhancements in the overall quadcopter performance, accompanied by reduced operational expenses. Simulations based on Coppelia V-rep are also presented to investigate the real-time application of the suggested system. The results demonstrate a decrease in liquid slosh amplitude and, consequently, a reduction in the control effort of the controller. These findings have significant implications for improving the quality of quadcopter control in various real-world applications. 30cm..

Place Hold on TRAJECTORY TRACKING FOR AGRICULTURAL DYNAMIC MULTI COPTER AERIAL ROBOT /

Disturbance Rejection and Roll Over Estimation for Control of Non-Linear Robotic System/

By Malik, Kamal Mazhar. . 216,p; 30,cm..

Place Hold on Disturbance Rejection and Roll Over Estimation for Control of Non-Linear Robotic System/

Novel Scalable Skeleton-Based Dynamic Sign Language Recognition by Learning Attention-Enhanced Efficient Spatio-Temporal Features/

By Naz, Neelma.

Place Hold on Novel Scalable Skeleton-Based Dynamic Sign Language Recognition by Learning Attention-Enhanced Efficient Spatio-Temporal Features/

A Machine Learning-Driven Approach to Multi-Variable Solar Panel Health Monitoring /

By Iqbal, Saeed . . 167p. , The health of solar PV systems is vital for maintaining efficiency, reliability, and safety. Although PV systems are inherently low-maintenance and lack moving parts, their performance depreciates over time due to environmental stressors such as dust accumulation, temperature fluctuations, microcracks, and panel misalignment. Effective monitoring systems enable the early detection of these issues, helping timely initiating the maintenance procedures, improving energy production, and extending system lifespan. This is of particular significance since improving PV efficiency through proactive management is often more cost-effective than expanding capacity through additional investments. This study presents an innovative approach for evaluating solar panel health using ML-driven analysis, as an alternate of traditional multi-sensor methodologies. The proposed technique extracts multiple attributes from a single data stream, employing ML algorithms to predict environmental impacts on PV performance. Sensor data was analysed to find key characteristics that accurately represent various aspects of panel health. Initially, temperature effects on a 100 kW grid-tied solar system deployed across various installation sites, including parking lots, rooftops, and ground-level arrays were examined. The results indicated that variations in operational temperature led to power losses of 27.95% in ground-mounted systems, 5.41% in rooftop systems, and 0.82% in parking-based systems. These outcomes were derived from temperature measurements at the three installation sites, further supported by weather station data. TMR sensors were also investigated for their ability to check multiple PV panel variables, particularly threeaxis orientation. Experimental findings revealed that the largest axis error values for Bx, By, and Bz, the magnetic field components along the X, Y, and Z axes, were below 0.1%, 0.3%, and 0.5%, respectively. However, TMR sensors proved insufficient for accurately capturing solar panel temperature fluctuations, prompting a shift toward EIS for a more comprehensive analysis. EIS experiments were conducted under controlled temperature conditions ranging from 10°C to 60°C, while maintaining all other environmental variables constant. The EIS response was recorded for 26 distinct parameters, providing a detailed dataset for analysis. The 26 distinct parameters are various electrical values obtained from the EIS experiments. ML classification of the EIS data using Caret-based techniques foundxvii the Light Gradient Boosting Machine and CatBoost Classifier as the more suitable models, achieving classification accuracies of 85.63% and 85.54%, respectively. Additionally, EIS was used to detect microcracks and dust accumulation under four real-world scenarios: (1) clean and healthy panels, (2) dusty but healthy panels, (3) clean panels with microcracks, and (4) dusty panels with microcracks. The Gradient Boost classifier achieved an impressive 97.5% accuracy in distinguishing these conditions. The experimental results indicated that combining EIS with machine learning provides a powerful approach for assessing solar panel temperature variations, detecting microcracks, and monitoring dust accumulation through DC power lines alone. This method demonstrates considerable potential for real-time PV system diagnostics, thereby enhancing predictive maintenance and operational monitoring. 30cm.

Place Hold on A Machine Learning-Driven Approach to Multi-Variable Solar Panel Health Monitoring /

AI Based Chromosomal Aberration Detection in Karyograms/

By Tabassum, Sumaira.

Place Hold on AI Based Chromosomal Aberration Detection in Karyograms/

A Machine Learning-Driven Approach to Multi-Variable Solar Panel Health Monitoring/

By Iqbal, Saeed.

Place Hold on A Machine Learning-Driven Approach to Multi-Variable Solar Panel Health Monitoring/

Automated Karyotyping: Segmentation and Classification /

By Umbreen, Neelam. . 130p. , Karyotyping continues to be the bedrock of cytogenetic diagnosis, providing key information on chromosomal abnormalities causative of a broad range of genetic disorders, developmental abnormalities, and cancers. But standard karyotyping is time-consuming, requires extensive specialist interpretation, and is vulnerable to human mistake and inefficiency, especially in high-volume clinical settings. Despite advances in medical image analysis, the automation of karyotyping faces persistent challenges, including the lack of large-scale annotated datasets, difficulties in segmenting overlapping chromosomes, and variability in chromosome morphology and staining. These challenges define a significant research gap in developing scalable, accurate, and clinically deployable deep learning models for automated chromosome analysis. In order to overcome this gap, we first present a large-scale, clinically annotated cytogenetic database, built from 1,311 patients and consisting of 10,057 karyograms with 514,949 manually annotated chromosome singlets. Furthermore, 3,935 metaphase images are annotated at the instance level in COCO format. This data set reflects true-world diversity, ranging from normal and abnormal karyotypes to different Giemsa (G-banding) staining intensities, structural abnormalities, and overlapping difficult cases, and thus presents a solid basis for the creation and testing of deep learning models within automated cytogenetics. Based on this work, we create two primary methodologies aimed at the fundamental tasks of karyotyping. For segmentation of chromosomes, we introduce a variant Mask R-CNN model involving an Attention-based Feature Pyramid Network (AttFPN), spatial attention, and a LastLevelMaxPool component to improve multi-scale feature representation and contextual perception. It enhances performance in difficult situations, including overlapping chromosomes and weak banding patterns, and gains considerable improvements in mean Average Precision (mAP) compared to standard baselines. For chromosome classification, we present the Dual Attention Multiscale Pyramid Network (DAMP), a specifically designed model that combines channel and spatial attention mechanisms to concentrate on discriminative features, as well as a multiscale pyramid architecture to cope with size, orientation, and quality variation in chromosomes. DAMP's highest classification accuracy is 96.76% on both public and commercial datasets, performingxv better than state-of-the-art models like ResNet-50, Vision Transformers, and Siamese Networks. Overall, this thesis provides interpretable and scalable deep learning models for automating chromosome classification and segmentation. Through the closure of key gaps in dataset quality, model resilience, and clinical utility, the work facilitates the insertion of clever decision-support systems into cytogenetic pipelines, ultimately leading to improved diagnostic reliability and efficiency in the face of chromosomal disorders. 30cm.

Place Hold on Automated Karyotyping: Segmentation and Classification /

Multimedia Analytics for Scene Content Understanding /

By ALI, HASNAIN . . 133p. , With the rapid expansion of video content, understanding how humans retain and recall visual data has become crucial. Memorability, a key neurocognitive process, plays a significant role in retaining and retrieving video content. While past research has explored image memorability, video memorability has received less attention, leaving a gap in robust computational models for predicting memorable video events. This thesis addresses this gap through a multi-phase study focused on video memorability prediction, scalable feature extraction, and behavior training for robotic systems. The first study introduces a novel framework that predicts episodic video memorability by fusing deep features, including text, color, and motion. Episodic sequences are generated using a Fuzzy FastText model and color histogram analysis, while scene objects are identified using a Faster Region-based Convolutional Neural Network (Faster R-CNN). The fusion of these features results in improved short- and long-term memorability, with a superior Spearman’s rank correlation of 0.6428 and 0.4285, respectively. The second study focuses on a robust Stacked Bin-Convolutional Neural Network (SB-CNN) and Sparse Low-Rank Regressor (SLRR) model. This model improves video event classification by employing a low-rank representation technique that reduces noise in video frames, leading to more accurate predictions. The Multi-Attribute Decision Making (MADM) technique is applied to enhance decision making, achieving a recall time of 49.9247 on public datasets. In the final study, a Trimmed Q-learning algorithm is introduced to optimize memorability-driven scene prediction in mobile robots. The training is conducted through online, short-term, and long-term learning modules, with significant improvements in memorability scores: 72.84% for short-term and online learning, and 68.63% for long-term learning. By linking these phases, this thesis presents an integrated framework that effectively addresses the challenges of video memorability prediction, robust feature scaling, and robotic decision-making, offering practical insights for both academic research and real-world applications. 30cm.

Place Hold on Multimedia Analytics for Scene Content Understanding /