Decomposition Level Research Articles

HEURISTIC-ASSISTED ADAPTIVE HYBRID DEEP LEARNING MODEL WITH FEATURE SELECTION FOR EPILEPSY DETECTION USING EEG SIGNALS

The word epilepsy is related to a neurological disease occurred by abnormalities of brain neurons. Timely detection of epilepsy is helpful for patients to decrease the mortality rate. To detect seizures, the Encephalogram (EEG) signals are analyzed based on monitoring the conditions of patients, and seizures can be detected from the EEG signal at appropriate times. The manual detection from the EEG signal requires more time for detecting the seizures and also it needs domain knowledge. The miss detection is eliminated by improving the classification performance in automatic epilepsy detection. Nowadays, deep learning models have not been greatly harnessed in the detection of epileptic seizures due to inappropriate descriptions of time-domain signals and sub-optimal classifier design. The aforementioned issues are combated by the novel Adaptive Hybrid Deep Learning (AHDL) approaches for epilepsy detection using EEG signals. Initially, the required EEG signal is collected from benchmark datasets. The collected signals are subjected to a signal decomposition phase that is accomplished by five levels of decomposition using Dual-Tree Complex Wavelet Transform (DTCWT), where the parameters are tuned by Improved Probability-based Coyote Optimization Algorithm (IP-COA). Further, the decomposed signal is given for feature extraction, where it divides the signal into two phases. In the first phase, the first feature set is obtained by using One-Dimensional Convolutional Neural Network (1DCNN), whereas in the second phase, the proposed model utilizes Auto Encoder (AE) to provide the second feature set. These resultant features are getting fused and the optimal feature selection process is found, where the features are obtained optimally by the IP-COA. Finally, epilepsy detection is accomplished with the aid of proposed AHDL with both Radial Basis-Recurrent Neural Networks (RB-RNN), where the hyperparameters are optimized using IP-COA. Thus, the experimental results illustrate that the suggested model enhances the detection and classification rate.

Faults Diagnosis for Large-Scale Battery Packs via Texture Analysis on Spatial–Temporal Images Converted From Electrical Behaviors

Battery failures have become the most intractable obstacles undermining the market confidence in applications like electric vehicle and power grid energy storage. This paper aims to fashion a generic diagnosis scheme against the faults in large-scale battery systems. First, a voltmeter array-based anomaly perception mechanism against the electrical behaviors of battery packs is developed. Then, system information on spatial arrangement and temporal dynamics is organically fused and drawn as a kind of pseudo 2D images (P2Is). Afterwards, by analyzing the resultant P2Is with the 2D variational mode decomposition (2D-VMD) and gray level co-occurrence matrix (GLCM), some statistical quantities concerning multi-scale texture features, extracted and refined by the principal component analysis (PCA), are found to have strong indicative associations with battery fault type and fault grade. Finally, relying on the multi-class relevance vector machine (M-RVM), feature evidences are synthesized to detect fault occurrence and give judgements on fault specifics of type and severity. Experimental verifications on a li-ion battery pack with 180 cells suggest that the proposed scheme behaves well in fault type isolating, with an accuracy rate of 97.6% and in fault severity grading, with an accuracy rate of 84.67%.

LW-BPNN: A Novel Feature Extraction Method for Rolling Bearing Fault Diagnosis

Efficiently diagnosing bearing faults is of paramount importance to enhance safety and reduce maintenance costs for rotating machinery. This paper introduces a novel bearing fault diagnosis method (LW-BPNN), which combines the rich properties of Legendre multiwavelet bases with the robust learning capabilities of a BP neural network (BPNN). The proposed method not only addresses the limitations of traditional deep networks, which rely on manual feature extraction and expert experience but also eliminates the complexity associated with designing and training deep network architectures. To be specific, only two statistical parameters, root mean square (RMS) and standard deviation (SD), are calculated on different Legendre multiwavelet decomposition levels to thoroughly represent more salient and comprehensive fault characteristics by using several scale and wavelet bases with various regularities. Then, the mapping relation between the extracted features and the health conditions of the bearing is automatically learned by the simpler BPNN classifier rather than the complex deep network structure. Finally, a few experiments on a popular bearing dataset are implemented to verify the effectiveness and robustness of the presented method. The experimental findings illustrate that the proposed method exhibits a high degree of precision in diagnosing various fault patterns. It outperforms other methods in terms of diagnostic accuracy, making it a viable and promising solution for real-world industrial applications in the field of rotating machinery.

Performance assessment of discrete wavelet transform for de-noising of FBG sensors signals embedded in asphalt pavement

In recent years, the Fiber Bragg Grating (FBG) sensor technology has been increasingly utilized as an optical measurement system in various engineering applications, particularly for structural health monitoring (SHM) purposes. This trend can be attributed to the inherent benefits of FBG sensors, such as their small size, immunity to electromagnetic interference, resistance to corrosion, and high accuracy and sensitivity. Various factors cause noise in the FBG sensor signal, which has a significant effect on measurement precision. As a result, de-noising plays an important role in the use of FBG sensor systems. In this study, strain data collected from FBG sensors embedded in a road section were used to evaluate the performance of discretized wavelet transform (DWT) for denoising FBG signals. The presence of noise poses a significant challenge in accurately measuring low-amplitude strains and light loads. To address this issue, various approaches have been investigated, including the selection of appropriate mother wavelets, levels of decomposition, thresholding functions, and thresholding selection approaches, with the aim of identifying the optimal parameters for effective denoising. The results show that FBG signals could be denoised successfully and low amplitude strains appeared completely without any loss of valuable data.

Multiresolution analysis for tool failure detection in CFRP/Ti6Al4V hybrid stacks drilling in aircraft assembly lines

This article presents a method for identifying Catastrophic Tool Failures during CFRP/Ti6Al4V hybrid stacks drilling in aircraft fuselage components assembly. The study is based on a real industrial case in the production system of a commercial aircraft section, where data was collected from the aircraft assembly line for 17 months. Over 30 different tool configurations were utilized in machining the aircraft section. However, the tool exhibiting the highest breakage rate, specifically a 4.6 mm diameter spiral fluted drill designed for CFRP/Ti6Al4V applications, was chosen for further analysis. This data collection included 69 samples of the selected cutting tool, revealing 19 instances of tool failure. The proposed detection method for spindle consumption signals utilizes Multiresolution Analysis (MRA) with Discrete Wavelet Transform (DWT) and Z-score technique. The research provides a comprehensive analysis of the use of DWT, explaining how the approximation and detail coefficients are obtained, which padding mode is suitable for the industrial case, the influence of the sliding process and filter bank length on the detection of sudden drops in the signal, and how to choose a suitable decomposition level in MRA based on the information provided by the scalograms. The confusion matrix results demonstrate the method's high performance, with Recall and Precision values equal to 1, which means that 100 % of tool failures were detected with no false positives. Additionally, the algorithm running time is 0.65 s, indicating that the method would apply to real-time tool failure detection in industrial machining production systems.

Improving Rock Classification with 1D Discrete Wavelet Transform Based on Laboratory Reflectance Spectra and Gaofen-5 Hyperspectral Data

The high intra-class variability of rock spectra is an important factor affecting classification accuracy. The discrete wavelet transform (DWT) can capture abrupt changes in the signal and obtain subtle differences between the spectra of different rocks. Taking laboratory spectra and hyperspectral data as examples, high-frequency features after DWT were used to improve the discrimination accuracy of rocks. Various decomposition levels, mother wavelet functions, and reconstruction methods were used to compare the accuracy. The intra-class variability was measured using the intra-class Spectral Angle Mapper (SAM). Our results show that the high-frequency features could improve the discrimination accuracy of laboratory spectra by 13.4% (from 46.5% to 59.9%), compared to the original spectral features. The accuracy of image spectra in two study areas increased by 8.6% (from 68.3% to 76.9%) and 7.2% (from 81.3% to 88.5%), respectively. Haar wavelets highlighted the spectral differences between different rocks. After DWT, intra-class SAM reduced and intra-class variability of rocks decreased. The Pearson correlation coefficient indicated a negative correlation between intra-class variability and overall accuracy. It suggested that improving classification accuracy by reducing intra-class variability was feasible. Though the result of lithological mapping still leaves room for improvement, this study provides a new approach to reduce intra-class variability, whether using laboratory spectra or hyperspectral data.

Abstract 15813: Machine Learning Improves Prediction of Heart Failure and Cardiovascular Death Through a Signal Analytical Approach of Echocardiography

Introduction: Recently, speckle tracking echocardiography (STE) and tissue Doppler imaging (TDI) have gained increasing traction as non-invasive tissue characterization methods within cardiology. But until now many patterns from the strain and TDI curves remain uninvestigated. Signal analytical methods like wavelet analysis have shown promising potential in effectively identifying previously unknown pathological patterns present in ECG signals. Therefore, we hypothesized that a signal analytic approach combined with supervised machine learning (ML) on strain and TDI curves could identify unknown pathophysiological deformation drivers of heart failure (HF) and cardiovascular death (CV death) in the general population. Methods: The analysis included 3781 subjects from the general population. In total 720 novel statistical parameters and wavelet signal parameters from 18 strain curves and 6 TDI curves were generated. The parameters were used to train an ensemble decision tree with 20-fold cross-validation and compared to a baseline ML model trained on 22 conventional echocardiographic parameters. Results: Follow up-time was four years. In total 108 subjects (2.9%) met the outcome. By including statistical and wavelet-derived parameters along with 22 conventional echocardiographic parameters in a combined model, the ML model was significantly improved compared to the baseline model (AUC for conventional model: 0.76 vs combined model: 0.825, p = 0.0086). Both statistical parameters, such as the mean and skewness of the TDI curve, as well as wavelet parameters, such as mean on the first decomposition level of the strain curve, were found to be important in the model. Conclusion Adding novel echocardiographic parameters based on signal analytical statistics and wavelet decomposition methods to preexisting conventional echocardiographic parameters significantly improved the prediction of incident HF or CV death in ML models.

Noise reduction and characteristic analysis of fluid signal in the jet impact‐negative pressure deamination reactor based on wavelet transform

AbstractThe jet impact‐negative pressure deamination reactor (JI‐NPDR) is a new type of continuous and efficient deamination equipment. The study of the random flow pattern of porous jet impingement in the reactor under negative pressure conditions is an important issue. In this work, the signal processing method based on wavelet transform is used to analyze the characteristics of random flow signals in the reactor. Meanwhile, an analog similar signal is built and three sets of Gaussian white noise with various signal‐to‐noise ratios are employed via the MATLAB platform. Based on the adjustment of threshold function, threshold, decomposition level and other parameters of wavelet transform, the noise ratio (SNR) and mean squared error (MSE) are used to evaluate the wavelet denoising effect. And then, the optimal denoising scheme for the obtained signal will be applied in processing the vacuum flow signal collected inside the deamination reactor. Subsequently, the 8‐layer wavelet decomposition is investigated by using sym7 as the wavelet basis, soft threshold function, and heursure threshold for signal denoising. Then, the analog signal is fed back through the results of the actual signal denoising, and the number of wavelet decomposition layers is adjusted from 8 to 9 layers to optimize the original wavelet denoising combination. By analyzing the low‐frequency and high‐frequency parts of the signal spectrum before and after denoising, it was found that wavelet transform can effectively denoise the fluid signal in the reactor.

Feature Selection Techniques for Wood Density Prediction in Forest Dataset

Feature selection becomes important, especially in data sets with a large number of variables and features. It will eliminate unimportant variables and increase classification accuracy and performance. Because of the increasing growth of data in numerous industries, some data is high-dimensional and contains essential and complex hidden linkages, posing new problems to feature selection: i) How to extract the underlying available relationships from the data, and ii) How to apply the learnt relations to improve feature selection? To address these issues, we use the six feature selection approach as a pre-processing step in the analysis to avoid over fitting and potential model under performance. Which can learn and apply the underlying sample relations and feature relations for feature selection. This study compared six feature selection approaches (Pearson Coefficient, Correlation matrix, Variable Importance, Forward selection, and Backward Elimination) for determining the decomposition level of forest trees. Our trials clearly provide a comparative evaluation of the Wrapper approach from several angles. Furthermore, we compare the dataset result with critical attributes to obtain the highest percentage accuracy. The experimental results show that the wrapper technique outperforms all other methods in all experiment groups.

Questioning the Impact of Vestibular Rehabilitation in Mal de Debarquement Syndrome

Introduction: Mal de debarquement syndrome (MdDS) is a rare and poorly understood clinical entity defined as a persistent sensation of rocking and swaying that can severely affect the quality of life. To date, the treatment options are very limited. Even though vestibular rehabilitation (VR) efficacy following peripheral vestibular lesion is well-documented, little is known about its influence on MdDS. The objective of the study was to explore the influence of traditional VR program on postural control in a patient diagnosed with MdDS. Methods: We assessed 3 different participants: 1 healthy control; 1 participant with identified peripheral vestibular impairment (VI); 1 participant diagnosed with MdDS. Postural control was assessed using a force plate (AMTI, Accusway). Participants were assessed following the modified Clinical Test Sensory Integration Balance protocol (mCTSIB, eyes open on firm surface/eyes closed on firm surface/eyes open on foam/eyes closed on foam). The raw data were exported and analyzed in a custom-made Matlab script (Matlab R2020a). We retrieved the center of pressure velocity in both anterior-posterior and mediolateral directions and performed an analysis of the frequency content using Daubechies wavelet of order 4 with 6 levels of decomposition. Protocol VI and MdDS patients performed a 4-week VR program. Postural control, using a force plate, and Dizziness Handicap Inventory (DHI) were assessed before and after the VR program. Healthy control was assessed twice separated by 1 week without any specific intervention. Results: VI participant showed clear improvement on DHI and sway velocity on condition eyes closed with foam. Accordingly, a reduction of energy content within frequency bands (0.39–0.78 Hz and 0.78–1.56 Hz) was observed post-rehabilitation for VI participant in both conditions with foam. Interestingly, MdDS participant demonstrated a reduction in sway velocity in most of the conditions but the frequency content was not modified by VR and was comparable to healthy control. Accordingly, the DHI of the MdDS participant failed to demonstrate any difference following VR. Conclusion: The results of the present study question the use of VR as an efficient treatment option for MdDS. Future studies must recruit a larger sample size and focus on the relationship between illusion of movement and postural characteristics such as sway velocity.

Investigation of noise suppression in experimental multi-cell battery string voltage applying various mother wavelets and decomposition levels in discrete wavelet transform for precise state-of-charge estimation

Lithium-ion battery state-of-charge (SOC) estimation algorithms developed employing electrical circuit models (ECMs) require original experimental discharging and charging voltage (EDCV) data for very high accuracy. However, if noisy EDCV signals are detected while sensing, inaccurate SOC estimation, which causes poor BMS performance, is unavoidable. Thus, this paper introduces a novel method for implementing noise suppression employing a discrete wavelet transform (DWT) multi-resolution analysis (MRA) based method. The approach extensively performs two comparison evaluations of denoising based on the information provided by the mother wavelet and the corresponding MRA level decomposition. The former approach is to obtain the level of decomposition with high execution of the denoising under a similar mother wavelet state. The second approach, on the other hand, compares noise suppression for various mother wavelets under identical conditions of the level decomposition with similar decomposition level conditions. Signal-to-noise ratio (SNR) is used to evaluate all comparative analyses. This method compares both hard- and soft-thresholding noise reduction methods in their entirety. From these findings, the study's analytical findings indicate a precise evaluation by demonstrating the variation in SNR with the corresponding mother wavelets and their decomposition levels. Thus, it is certain that the mother wavelet and the decomposition level are optimized for noise suppression for a battery string, series/parallel (2S3P) each of 2.2 Ah, with the EDCV signal incorporated with noise.

Development of CNN Model to Avoid Food Spoiling Level

Food spoilage is a pervasive issue that contributes to food waste and poses significant economic and environmental challenges worldwide. To combat this problem, we propose the development of a Convolutional Neural Network (CNN) model capable of predicting and preventing food spoilage. This paper outlines the methodology, data collection, model architecture, and evaluation of our CNN-based solution, which aims to assist consumers, retailers, and food producers in minimizing food waste. Researchers are working on innovative techniques to preserve the quality of food in an effort to extend its shelf life since grains are prone to spoiling as a result of precipitation, humidity, temperature, and a number of other factors. In order to maintain current standards of food quality, effective surveillance systems for food deterioration are needed. To monitor food quality and control home storage systems, we have created a prototype. To start, we used a Convolutional Neural Network (CNN) model to identify the different types of fruits and vegetables. The suggested system then uses sensors and actuators to check the amount of food spoiling by monitoring the gas emission level, humidity level, and temperature of fruits and vegetables. Additionally, this would regulate the environment and, to the greatest extent feasible, prevent food spoiling. Additionally, based on the freshness and condition of the food, a message alerting the client to the food decomposition level is delivered to their registered cell numbers. The model used turned out to have a 96.3% accuracy rate.

A Novel and Optimized Sine–Cosine Transform Wavelet Threshold Denoising Method Based on the sym4 Basis Function and Adaptive Threshold Related to Noise Intensity

In digital shearography, the speckle noise of the phase fringe pattern has a negative effect on the accuracy and reliability of the phase unwrapping procedure. A novel and optimized sine–cosine transform wavelet threshold denoising method is proposed to suppress speckle noise. Fast phase denoising can be achieved by using the proposed method while preserving the phase reversal information. The details of the selected wavelet basis function, the optimal decomposition level, the threshold function, and the denoising threshold are also provided in this manuscript. In particular, the decomposition level is analyzed and optimized through simulation analysis according to the speckle suppression index and the adaptive denoising method. The experimental results show that the proposed method has more adaptive ability in practical application than the sine–cosine transform average denoising method with the selected mask and iterative procedure, which speeds the denoising process up and takes better-unwrapped phase patterns.

Estimation of leaf nitrogen content in winter wheat based on continuum removal and discrete wavelet transform

ABSTRACT Accurate and rapid estimation of leaf nitrogen content (LNC) in winter wheat using hyperspectral techniques is important for growth monitoring and accurate fertilization. Based on field experiments conducted over four years at three ecological sites with four different nitrogen (N) application treatments and four different N-efficient wheat varieties in Henan Province, the canopy spectra and LNC of winter wheat were obtained simultaneously in the main growth stages. The original spectrum was subjected to continuum removal (CR), the correlation between LNC and various vegetation indices was systematically analysed, and the optimal vegetation index estimation model for LNC was constructed. Simultaneously, discrete wavelet transform (DWT) was used to further compress and extract the CR spectrum. The model was combined with partial least squares regression (PLSR) and K-nearest neighbour (KNN) algorithms to estimate LNC in winter wheat. The results showed that the spectrum treated with CR was significantly improved in its correlation with LNC and that the model established based on normalized vegetation index NDVI (CR728, CR977) combined with the CR spectrum was better compared to existing vegetation indices. The calibration and validation coefficients of determination (R2) and root mean square error (RMSE) were 0.816 and 0.799, and 0.352% and 0.342%, respectively. DWT was used to transform the CR spectrum, and the results showed that a combination of the CR spectrum and a sym8 wavelet function with PLSR based on the approximation coefficients at decomposition level 2 predicted LNC most accurately. The coefficient of determination RC 2, root mean square error RMSEC, and relative percent deviation (RPDC) of the calibration set were 0.884, 0.279%, and 2.932, respectively, and RV 2, RMSEV, and RPDV of the validation set were 0.855, 0.291%, and 2.619, respectively, indicating that the model had good stability and predictive ability. Combination of CR and DWT improved the modelling accuracy of wheat LNC and showed better prediction results for multi-year and multi-point samples. The results of the study can provide a basis and reference for rapid monitoring of LNC in winter wheat.

A Novel Methodology for Fault Sensing in Presence of Power Swing for Transmission Line with Thyristor Controlled Series Compensator

Digital distance relaying is implemented in extra high voltage (EHV) transmission network for speedy and reliable fault detection. But it should not activate in case of slow transients known as power swings. However, for occurrence of any type of fault during power swing, the relay should sense the faulty condition and send trip command to concerned circuit breaker. The detection of fault in a transmission line network with TCSC during the power swing condition has become further complex due to transients generated by series capacitor and the metal–oxide varistor (MOV) protecting it. This paper firstly presents effect of TCSC on distance relay operation for varied TCSC degree of compensation and firing angle values. Further, a discrete wavelet transform-based fast acting fault and swing classification algorithm is proposed which can sense all types of faults during slow as well as fast power swing in first decomposition level and within 0.001 sec. The proposed utilizes optimized threshold values both for swing and fault sensing by using Honey Bee Optimization Algorithm (HBOA). The proposed novel algorithm is coded in MATLAB software and the test system comprising of 400kV, 50Hz parallel transmission line network along with TCSC is built using MATLAB Simulink environment with sim power systems toolbox. It is tested for high/low resistance faults, symmetrical/asymmetrical faults and close-in/far end faults by changing TCSC compensation level and firing angle value. The simulation results prove the accuracy of proposed methodology for swing and fault classification.

Structural and Functional Characteristics of Soil Fungal Communities near Decomposing Moso Bamboo Stumps

Background and Objectives: Fungi degrade lignin and other fibers, thus playing an essential role in the decomposition of Phyllostachys edulis (Carrière) J.Houz. (Moso bamboo) stumps. Herein, we characterized key soil fungal communities near different levels of decomposing Moso bamboo stumps (mildly, moderately, and heavily decayed). Materials and Methods: High-throughput sequencing technology was used to analyze the soil fungal communities inside and outside of mild, moderate, and heavy decomposing Moso bamboo stumps. Results: We found nine phyla, 30 classes, 77 orders, 149 families, and 247 genera of soil fungi near the bamboo stumps. Soil fungi OTUs and diversity and richness indices were lower outside than inside the stumps, and decreased with increasing degrees of decay. Inside the bamboo stumps, Soil fungi OTUs and diversity and richness indices were the highest and lowest in moderate and heavy decay bamboo stumps, respectively. Ascomycota dominated inside (from 81% to 46%) and outside (from 69% to 49%) the stumps, and their relative abundance gradually decreased with decomposition, whereas that of Basidiomycota increased outside the stumps (from 17% to 49%). Two-way ANOVA showed that the interaction between the two factors of occurring inside and outside the bamboo stumps and the degree of decay, significantly affected Chytridiomycota and Penicillium (p < 0.001) and significantly affected Mucoromycota (p < 0.05). The abundance of different genera was significantly correlated with saprotrophic functional groups. Conclusion: Changes in the structure and functional groups of soil fungal communities may play an important role during different levels of decomposition of Moso bamboo stumps. This study provides a scientific basis for screening functional fungal strains that promote the decomposition of Moso bamboo stumps.

The Fungal Development During The Leaf Litter Process Decomposition of Avicennia marina

The existence of mangrove ecosystems decreases from land conversion into residential areas, industries, plantations, road facilities and the construction of ponds. This research was conducted in the mangrove forest area of Secanang Belawan Village, Medan, at the Biotechnology Laboratory of the Department of Forestry, Faculty of Agriculture, USU and at the Microbiology Laboratory, Department of Biology, Faculty of Mathematics and Natural Sciences, USU Medan. The study's objective was to identify the different fungal species that were present at 10 ppt salinity levels, 10–20 ppt, and 20–30 ppt as well as the pace at which Avicennia marina leaf litter. The research method used litter bags filled with leaf litter of Avicennia marina (50 g) with 7 treatments, 3 replications and 3 levels of salinity. The A. marina leaf litter that had a salt level of 10 ppt had the highest level of decomposition. The dry weight of leaf litter left in the litter bag, which is 4.92 g on average. The largest dry weight of the remaining litter was found in an environment with a salinity of 20 – 30 ppt, which is an average of 10.42 g. The rate of decomposition of A. marina leaf litter in an environment with a salt level greater than 30 ppt is 6.53/yr (or almost equal to residence period of = 0.15 yr). The number of fungal species discovered in the leaf litter of A. marina, which goes through a breakdown process at salinity levels between 0 and 10 ppt and contains 9 species, is another way to observe the impact of salt level. In comparison to the large at salinity levels of 10 to 20 ppt and 20 to 30 ppt, respectively, there were populations of fungus that underwent the process of decomposition in A. marina leaf litter, with an average of 5.99 x 102 cfu/ml and 5.5 x 102 cfu/ml, the fungal population that was the highest concentration was 10.72 x 102 cfu/ml

A new criteria for determining the best decomposition level and filter for wavelet-based data-driven forecasting frameworks- validating using three case studies on the CAMELS dataset

A new criteria for determining the best decomposition level and filter for wavelet-based data-driven forecasting frameworks- validating using three case studies on the CAMELS dataset

Aquaculture defects recognition via multi-scale semantic segmentation

Aquaculture net pen defects such as biofouling, vegetation, and holes are key challenges to efficient and sustainable fish production in aquaculture. These defects must be monitored to avoid problems with aquaculture net safety as well as fish growth. Recently, deep learning methods have been adopted to solve detection and classification problems in different applications. However, the conventional methods are challenging to meet the demands of high precision and real-time detection of aquaculture net defects in a complex marine environment. Towards this end, this paper proposes an autonomous net pen defect detection system that contains a novel multi-scale semantic segmentation topology for detecting the biofouling, vegetation, and hole problems in the aquaculture environment. In particular, we emphasize fusing the attention maps obtained across different decomposition levels of the network to generate rich feature distributions that enable the accurate extraction of biofouling, vegetation, and holes within aquaculture nets. Moreover, the proposed model is thoroughly tested on two private datasets and two public datasets which were acquired from a real-field fish farms. Across all four datasets, the proposed framework showed remarkable biofouling, vegetation, and hole detection performance, where it outperformed state-of-the-art methods by 6.58%, 3.69%, 6.44%, and 4.78% in terms of mean average precision across LABUST, KU, NDv1, and NDv2 datasets, respectively.

Application of Hybrid ANN Techniques for Drought Forecasting in the Semi-Arid Region of India.

The intensity and frequency of diverse hydro-meteorological disasters viz., extreme droughts, severe floods, and cyclones have increasing trends due to unsustainable management of land and water resources, coupled with increasing industrialization, urbanization and climate change. This study focuses on the forecasting of drought using selected Artificial Neural Network (ANN)-based models to enable decision-makers to improve regional water management plans and disaster mitigation/reduction plans. Four ANN models were developed in this study, viz., one conventional ANN model and three hybrid ANN models: (a) Wavelet based-ANN (WANN), (b) Bootstrap based-ANN (BANN), and (c) Wavelet-Bootstrap based-ANN (WBANN). The Standardized Precipitation Evapotranspiration Index (SPEI), the best drought index identified for the study area, was used as a variable for drought forecasting. Three drought indices, such as SPEI-3, SPEI-6 and SPEI-12 respectively representing "short-term", "intermediate-term", and "long-term" drought conditions, were forecasted for 1-month to 3-month lead times for six weather stations over the study area. Both statistical and graphical indicators were considered to assess the performance of the developed models. For the hybrid wavelet model, the performance was evaluated for different vanishing moments of Daubechies wavelets and decomposition levels. The best-performing bootstrap-based model was further used for analysing the uncertainty associated with different drought forecasts. Among the models developed for drought forecasting for 1 to 3 months, the performances of the WANN and WBANN models are superior to the simple ANN and BANN models for the SPEI-3, SPEI-6, and SPEI-12 up to the 3-month lead time. The performance of the WANN and WBANN models is the best for SPEI-12 (MAE = 0.091-0.347, NSE = 0.873-0.982) followed by SPEI-6 (MAE = 0.258-0.593; NSE = 0.487-0.848) and SPEI-3 (MAE = 0.332-0.787, NSE = 0.196-0.825) for all the stations up to 3-month lead time. This finding is supported by the WBANN analyze uncertainties as narrower band width for SPEI-12 (0.240-0.898) as compared to SPEI-6 (0.402-1.62) and SPEI-3 (0.474-2.304). Therefore, the WBANN model is recommended for the early warning of drought events as it facilitates the uncertainty analysis of drought forecasting results.