Self-adaptive Threshold Research Articles

Adaptive Threshold Algorithm for Outlier Elimination in 3D Topography Data of Metal Additive Manufactured Surfaces Obtained from Focus Variation Microscopy

The topography of surfaces produced by metal additive manufacturing is a challenge for optical measurement systems such as focus variation microscopes. These irregularities can lead to artifacts, such as incorrectly measured protrusions or spikes, hampering reliable topographic characterization. In order to eliminate this problem, we introduce a new algorithm based on dual convolving a vertical Sobel operator with cross sections of an image stack parallel to the scanning direction of the so-called depth scan. This has proven beneficial in order to distinguish the focus region from out-of-focus areas where outliers are frequently detected. This paper introduces a method for deriving self-adaptive thresholds from the convolution result and compares the effects of different operators in creating self-adaptive thresholds. Additionally, a simulation model of focus variation microscopy is introduced to validate both the measuring system and the proposed algorithm, thereby enhancing the overall performance of focus variation microscopy. Finally, comparisons of measurement results on rough metal additive manufacturing workpieces with and without self-adaptive thresholds are discussed to demonstrate the algorithm’s effectiveness.The utilization of self-adaptive thresholds demonstrably reduces the uncertainty range in roughness parameter calculations. For example, in the case of an additive manufactured metal sample due to outlier elimination, the Sz roughness value reduces from 543 µm to 413 µm.

Enhancing 3D object detection by using neural network with self-adaptive thresholding

Robust 3D object detection remains a pivotal concern in the domain of autonomous field robotics. Despite notable enhancements in detection accuracy across standard datasets, real-world urban environments, characterized by their unstructured and dynamic nature, frequently precipitate an elevated incidence of false positives, thereby undermining the reliability of existing detection paradigms. In this context, our study introduces an advanced post-processing algorithm that modulates detection thresholds dynamically relative to the distance from the ego object. Traditional perception systems typically utilize a uniform threshold, which often leads to decreased efficacy in detecting distant objects. In contrast, our proposed methodology employs a Neural Network with a self-adaptive thresholding mechanism that significantly attenuates false negatives while concurrently diminishing false positives, particularly in complex urban settings. Empirical results substantiate that our algorithm not only augments the performance of 3D object detection models in diverse urban and adverse weather scenarios but also establishes a new benchmark for adaptive thresholding techniques in field robotics.

Enhancing 3D object detection by using neural network with self-adaptive thresholding

Robust 3D object detection remains a pivotal concern in the domain of autonomous field robotics. Despite notable enhancements in detection accuracy across standard datasets, real-world urban environments, characterized by their unstructured and dynamic nature, frequently precipitate an elevated incidence of false positives, thereby undermining the reliability of existing detection paradigms. In this context, our study introduces an advanced post-processing algorithm that modulates detection thresholds dynamically relative to the distance from the ego object. Traditional perception systems typically utilize a uniform threshold, which often leads to decreased efficacy in detecting distant objects. In contrast, our proposed methodology employs a Neural Network with a self-adaptive thresholding mechanism that significantly attenuates false negatives while concurrently diminishing false positives, particularly in complex urban settings. Empirical results substantiate that our algorithm not only augments the performance of 3D object detection models in diverse urban and adverse weather scenarios but also establishes a new benchmark for adaptive thresholding techniques in field robotics.

Hierarchical Context Representation and Self-Adaptive Thresholding for Multivariate Anomaly Detection

Hierarchical Context Representation and Self-Adaptive Thresholding for Multivariate Anomaly Detection

A holistic semi-supervised method for imbalanced fault diagnosis of rotational machinery with out-of-distribution samples

Fault diagnosis plays a critical role in ensuring the reliability and safety of industrial systems. Despite the success of semi-supervised learning in fault diagnosis, challenges persist in dealing with imbalanced datasets and out-of-distribution (OOD) samples, particularly in industrial settings where fault incidents are rare. The rarity of fault incidents in industrial contexts leads to dataset imbalances, while changes in conditions may introduce OOD samples. To address these challenges, this study proposes a novel semi-supervised fault diagnosis method that simultaneously considers the imbalanced distribution of fault categories and OOD samples. Firstly, the semi-supervised framework has been presented, which combines pseudo-labeling and consistency regularization techniques. Then, the self-adaptive threshold has been devised to balance confidence levels among fault classes using dataset-specific and class-specific thresholds, thereby enhancing the model's robustness in real-world scenarios. Additionally, the OOD detection strategy based on contrastive learning has been proposed, wherein predictions from two different views of each sample estimate its “likelihood” of belonging to in-distribution (ID) or OOD categories. This strategy improves system safety by alerting operators to potential anomalies. Comprehensive experimental validations and comparisons were conducted on two rotational machinery fault datasets, demonstrating that the proposed method achieves superior performance compared to the alternative methods and is applicable for practical application.

Validation and refinement of cropland map in southwestern China by harnessing ten contemporary datasets

Accurate cropland map serves as the cornerstone of effective agricultural monitoring. Despite the continuous enrichment of remotely sensed cropland maps, pervasive inconsistencies have impeded their further application. This issue is particularly evident in areas with limited valid observations, such as southwestern China, which is characterized by its complex topography and fragmented parcels. In this study, we constructed multi-sourced samples independent of the data producers, taking advantage of open-source validation datasets and sampling to rectify the accuracy of ten contemporary cropland maps in southwestern China, decoded their inconsistencies, and generated a refined cropland map (CroplandSyn) by leveraging ten state-of-the-art remotely sensed cropland maps released from 2021 onwards using the self-adaptive threshold method. Validations, conducted at both prefecture and county scales, underscored the superiority of the refined cropland map, aligning more closely with national land survey data. The refined cropland map and samples are publicly available to users. Our study offers valuable insights for improving agricultural practices and land management in under-monitored areas by providing high-quality cropland maps and validation datasets.

New cardiovascular disease prediction approach using support vector machine and quantum-behaved particle swarm optimization

Cardiovascular disease (CVD) is one of the leading causes of death worldwide. Early detection of CVD reduces the risk of a heart attack and increases the chance of recovery. The use of angiography to detect CVD is expensive and has negative side effects. In addition, existing CVD diagnostic methods usually achieve low detection rates and reach the best decision after many iterations with low convergence speeds. Therefore, a novel heart disease detection model based on the quantum-behaved particle swarm optimization (QPSO) algorithm and support vector machine (SVM) classification model, namely, QPSO-SVM, was proposed to analyze and predict heart disease risk. First, the data preprocessing was performed by transforming nominal data into numerical data and applying effective scaling techniques. Next, the SVM fitness equation is expressed as an optimization problem and solved using the QPSO to determine the optimal features. Finally, a self-adaptive threshold method for tuning the QPSO-SVM parameters is proposed, which permits it to drop into local minima, and balances between exploration and exploitation in the solution search space. The proposed model is applied to the Cleveland heart disease dataset and compared with state-of-the-art models. The experimental results show that the proposed QPSO-SVM model achieved the best heart-disease-prediction accuracies of 96.31% on the Cleveland heart data set. Furthermore, QPSO-SVM outperforms other state-of-the-art prediction models considered in this research in terms of sensitivity (96.13%), specificity (93.56%), precision (94.23%), and F1 score (0.95%).

Fault diagnosis in bipolar HVDC systems based on traveling wave theory by monitoring data from one terminal

This paper proposes a method for detecting, classifying and locating incident faults on transmission line (TL) of a bipolar HVDC system. This method is based on the traveling waves theory and uses the redundant discrete wavelet transform to filter voltage signals monitored at only one system terminal. The model represents the Brazilian HVDC system of the Madeira River, which is simulated in the Alternative Transient Program (ATP). Different fault scenarios are analyzed, being generated by varying type, resistance and fault location. Moreover, in order to make the simulation more realistic, the TL is modeled based on frequency dependent parameters. Thereby, the impacts of line parameter uncertainties and transient attenuation on the proposed method are analyzed. From the obtained results, it is concluded that the method correctly detects all fault types through the use of a self-adaptive detection threshold. Furthermore, it is verified that the self-adaptive threshold detects the fault times instants with an efficiency superior to the fixed threshold. In the classification step, all faults were correctly classified through rules elaborated based on the voltage variation level. Finally, the total average errors of fault location represents only 0.047% of the TL extension. Moreover, it is verified that fault location errors for pole:pole faults were inferior to those obtained for pole:ground faults, being the efficiency of the method inversely proportional to the fault resistance.

Multi-sensor imaging of winter buried lakes in the Greenland Ice Sheet

Recent studies have highlighted that meltwater in supraglacial lakes (SLs) can be buried during frozen season in the Greenland Ice Sheet (GrIS). Meltwater in buried lakes (BLs) can even persist through the winter, disturbing the englacial thermal regime and providing an important buffer against GrIS's contribution to sea-level rise. However, little is known about the inter-annual BL dynamics in the GrIS, and there is no quantitative statistic about the overall buried percentage. Here, we conduct a satellite-based study to automatically map the winter BLs over the GrIS during 2017–2022 using multi-source optical and synthetic aperture radar (SAR) images on the Google Earth Engine (GEE) platform. To eliminate the interferences from other weak microwave reflecting surfaces, summer SLs are first extracted from Landsat 8 and Sentinel-2 images to determine the potential BL searching areas on winter Sentinel-1 images. A self-adaptive thresholding algorithm is proposed to extract BLs within the dilated summer SLs using histogram-based morphological edge detectors. BLs extracted by the proposed method and visual interpretation show a substantial agreement with a precision of 0.82 and a Kappa coefficient of 0.70. On average, a total buried lake area of 182.27 km2 was observed each winter during the period 2017–2022. BLs were mainly distributed in the Center-West, South-West and North-East Basins, with the majority occurring at elevations between 800 and 1700 m. In 2019–2020, a sudden extension of BLs was observed over the GrIS, especially in the North-East Basin where abnormally high temperatures and surface runoff were recorded. In 2021–2022, a widespread distribution of BLs in the South-West Basin was observed after abnormal snowfall. Overall, about 13% of the GrIS summer SLs can persist through winter, suggesting the potential for meltwater hydrofracture in winter over large areas.

A Self-Adaptive Thresholding Approach for Automatic Water Extraction Using Sentinel-1 SAR Imagery Based on OTSU Algorithm and Distance Block

As an indispensable material for animals, plants and human beings, obtaining accurate water body information rapidly is of great significance to maintain the balance of ecosystems and ensure normal production and the life of human beings. Due to its independence of the time of day and the weather conditions, synthetic aperture radar (SAR) data have been increasingly applied in the extraction of water bodies. However, there is a great deal of speckle noise in SAR images, which seriously affect the extraction accuracy of water. At present, most of the processing methods are filtering methods, which will cause the loss of detailed information. Based on the characteristic of side-looking SAR, this paper proposed a self-adaptive thresholding approach for automatic water extraction based on an OTSU algorithm and distance block. In this method, the whole images were firstly divided into uniform image blocks through a distance layer which was produced by the distance to the orbit. Then, a self-adaptive processing was conducted for merging blocks. The OTSU algorithm was used to obtain a threshold for classification and the Jeffries–Matusita (JM) distance was calculated with the classification result. The merge processing continued until the separability of image blocks reached the maximum. Subsequently, we started from the next block to repeat the merger, and so on until all blocks were processed. Ten study areas around the world and the local Dongting Lake area were applied to test the feasibility of the proposed method. In comparison with five other global threshold segmentation algorithms such as the traditional OTSU, MOMENTS, MEAN, ISODATA and MINERROR, the proposed method obtains the highest overall accuracy (OA) and kappa coefficient (KC), while this approach also demonstrates greater robustness in the analysis of time series. The findings of this study offer an effective method to improve water detection accuracy as well as reducing the influence of speckle noise and retaining details in the image.

Detection of Arrhythmia Heartbeats from ECG Signal Using Wavelet Transform-Based CNN Model

In India, over 25,000 people have died from cardiovascular annually over the past 4 years , and over 28,000 in the previous 3 years. Most of the deaths nowadays are mainly due to cardiovascular diseases (CVD). Arrhythmia is the leading cause of cardiovascular mortality. Arrhythmia is a condition in which the heartbeat is abnormally fast or slow. The current detection method for diseases is analyzing by the electrocardiogram (ECG), a medical monitoring technique that records heart activity. Since actuations in ECG signals are so slight that they cannot be seen by the human eye, the identification of cardiac arrhythmias is one of the most difficult undertakings. Unfortunately, it takes a lot of medical time and money to find professionals to examine a large amount of ECG data . As a result, machine learning-based methods have become increasingly prevalent for recognizing ECG features. In this work, we classify five different heartbeats using the MIT-BIH arrhythmia database . Wavelet self-adaptive thresholding methods are used to first denoise the ECG signal. Then, an efficient 12-layer deep 1D Convolutional Neural Network (CNN) is introduced for better features extraction, and finally, SoftMax and machine learning classifiers are applied to classify the heartbeats. The proposed method achieved an average accuracy of 99.40%, precision of 98.78%, recall of 98.78%, and F1 score of 98.74%, which clearly show that it outperforms with the exiting model . Architecture of proposed work is simple but effective in remote cardiac diagnosis paradigm that can be implemented on e-health devices.

A Novel ST-ViBe Algorithm for Satellite Fog Detection at Dawn and Dusk

Satellite remote sensing provides a potential technology for detecting fog at dawn and dusk on a large scale. However, the spectral characteristics of fog at dawn and dusk are similar to those of the ground surface, which makes satellite-based fog detection difficult. With the aid of time-series datasets from the Himawari-8 (H8)/AHI, this study proposed a novel algorithm of the self-adaptive threshold of visual background extractor (ST-ViBe) model for satellite fog detection at dawn and dusk. Methodologically, the background model was first built using the difference between MIR and TIR (BTD) and the local binary similarity patterns (LBSP) operator. Second, BTD and scale invariant local ternary pattern (SILTP) texture features were coupled to form scene factors, and the detection threshold of each pixel was determined adaptively to eliminate the influence of the solar zenith angles. The background model was updated rapidly by accelerating the updating rate and increasing the updating quantity. Finally, the residual clouds were removed with the traditional cloud removal method to achieve accurate detection of fog at dawn and dusk over a large area. The validation results demonstrated that the ST-ViBe algorithm could detect fog at dawn and dusk precisely, and on a large scale. The probability of detection, false alarm ratio, and critical success index were 72.5%, 18.5%, 62.4% at dawn (8:00) and 70.6%, 33.6%, 52.3% at dusk (17:00), respectively. Meanwhile, the algorithm mitigated the limitations of the traditional algorithms, such as illumination mutation, missing detection, and residual shadow. The results of this study could guide satellite fog detection at dawn and dusk and improve the detection of similar targets.

A fully-mapped and energy-efficient FPGA accelerator for dual-function AI-based analysis of ECG.

In this paper, a fully-mapped field programmable gate array (FPGA) accelerator is proposed for artificial intelligence (AI)-based analysis of electrocardiogram (ECG). It consists of a fully-mapped 1-D convolutional neural network (CNN) and a fully-mapped heart rate estimator, which constitute a complementary dual-function analysis. The fully-mapped design projects each layer of the 1-D CNN to a hardware module on an Intel Cyclone V FPGA, and a virtual flatten layer is proposed to effectively bridge the feature extraction layers and fully-connected layer. Also, the fully-mapped design maximizes computational parallelism to accelerate CNN inference. For the fully-mapped heart rate estimator, it performs pipelined transformations, self-adaptive threshold calculation, and heartbeat count on the FPGA, without multiplexed usage of hardware resources. Furthermore, heart rate calculation is elaborately analyzed and optimized to remove division and acceleration, resulting in an efficient method suitable for hardware implementation. According to our experiments on 1-D CNN, the accelerator can achieve 43.08× and 8.38× speedup compared with the software implementations on ARM-Cortex A53 quad-core processor and Intel Core i7-8700 CPU, respectively. For the heart rate estimator, the hardware implementations are 25.48× and 1.55× faster than the software implementations on the two aforementioned platforms. Surprisingly, the accelerator achieves an energy efficiency of 63.48 GOPS/W, which obviously surpasses existing studies. Considering its power consumption is only 67.74mW, it may be more suitable for resource-limited applications, such as wearable and portable devices for ECG monitoring.

Application of FY-4B Geostationary Meteorological Satellite in Grassland Fire Dynamic Monitoring

In this study, the channel data related to fire point identification of Advanced Geostationary Radiation Imager on Fengyun-4B (FY-4B/AGRI) and Advanced Meteorological Imager on GEO-KOMPSAT 2A (GK-2A/AMI) were cross-compared. A total of 267 sampling points in China (Guangdong, Guangxi, Guizhou, Yunnan, Hainan) were selected to carry out fine positioning correction on the data in different elevation intervals. Then, a fire monitoring algorithm based on FY-4B was proposed, in which the self-adaptive threshold adjustment of the underlying surface parameters and the reprocessing module of fire point identification were added. The algorithm can realize the high-precision and stable monitoring of fire. The continuous dynamic monitoring was carried out using the grassland fire Mongolia from April 18 to 20, 2022 as an example. The results showed that the parameters of FY-4B/AGRI and GK2A/AMI channels have high consistency. The root mean square error (RMSE) of reflection channel was 1.33%, and the maximum RMSE of brightness temperature channel was less than 1.3 Kelvin (K). Through the positioning analysis in different elevation intervals, the average longitude offset of FY-4B satellite data was -0.5°–0° and the average latitude offset was -0.9°–0.6°. Overall, these findings indicate that the high-frequency observations of FY-4B can be fully utilized to monitor forest and grassland fires, which can continuously track the dynamic evolution of fire and can distinguish the spatial distribution of different fire intensities in large-scale fire field.

Concurrent topology and fiber orientation optimization method for fiber-reinforced composites based on composite additive manufacturing

Additive manufacturing of fiber-reinforced composite (FRC) parts provides a new rapid prototyping technique for designs. However, restricted by the layer-by-layer style of additive manufacturing techniques, the designed fiber orientations should be parallel to the print plane, and that will result in a reduced design space for the orientation optimization. To enlarge the orientation design space under this constraint, this paper proposes a new optimization method for FRC parts. The optimized structure is composed of several groups of components, and the fiber orientation of each group of components is constrained to a specific plane. Based on the composite additive manufacturing technology, these components can be fabricated separately and joined together to form the whole structure. The number of the component groups (also the fiber print planes) is predefined. In this method, the structure is optimized by concurrently designing the topology, the components partitioning and the fiber orientations. The structural topology and components partitioning are described by the SIMP interpolation scheme and discrete material optimization (DMO) interpolation scheme, respectively. The discrete-continuous parameterization (DCP) method with two sub-intervals is applied in the orientation optimization to avoid the local optimum of the angle variables, and a unified formulation is given for the concurrent optimization design. Furthermore, a density projection strategy with self-adaptive threshold parameters for multi-material optimization problems is proposed to get the distinct component interfaces. A stable convergence can be obtained without manually adjusting the projection parameters during the optimization iterations. The corresponding angle filtering method is developed to smooth the optimized fiber orientations. Numerical examples and manufacturing experiments are presented to demonstrate the effectiveness of the proposed method.

Threshold optimization method of single-bit quanta image sensor based on bit density

To improve the quality of reconstructed images of single-bit quanta image sensors, a self-adaptive threshold optimization method is proposed. The method can adaptively adjust the quantization threshold according to different light intensities, making the quality of the reconstructed image improved. First, the bitstream data from the quantization of real images are obtained by sensor model. Second, the optimal threshold value of each row is determined by calculating the bit-density, followed by the threshold value updated according to the deviation value between the bit-density of a single-pixel and 0.5. Finally, the optimal threshold matrix obtained by this method is used to reconstruct the image to improve imaging quality. Four identical test images were processed by the proposed method, the traditional method, and Elgendy’s variable threshold method, respectively. The results show that the proposed method can improve the peak signal-to-noise ratio of the reconstructed image in comparison with the traditional uniform threshold method. Compared with Elgendy’s method, the proposed method can effectively reduce the number of iterations under the same parameter conditions. These experiments prove that the proposed method can effectively improve the quality of reconstructed images with fewer iterations.

Weak fault feature extraction of rolling bearings based on improved ensemble noise-reconstructed EMD and adaptive threshold denoising

Extracting weak fault features under noise interference is crucial for the fault diagnosis of rolling bearings at an early stage. In this paper, a new method based on improved ensemble noise-reconstructed empirical mode decomposition (IENEMD) and adaptive threshold denoising (ATD) is proposed for the weak fault feature extraction of rolling bearings. Firstly, to tackle the drawbacks of EEMD which utilizes the additional Gaussian white noise resulting in submersion of the weak fault features, the inherent noise hidden in the raw signals is automatically extracted and leveraged in the IENEMD to decompose the raw signals into intrinsic mode functions (IMFs). The IENEMD not only avoids the mode mixing issues of the original EMD but also reduces the interference of inherent noise. Then, the ATD consisting of informative IMF selection and threshold denoising is executed on the decomposed IMFs. Taking the health signals of rolling bearings as the benchmarks, the meaningful IMFs rich in fault information are efficiently selected, which are further denoised by a newly constructed self-adaptive threshold. Finally, weak fault features are extracted from the reconstructed denoised signals employing the envelope analysis approach. A simulation case and two actual cases of rolling bearings in the early fault stage are utilized to verify the robustness and feasibility of the proposed IENEMD-ATD. The results indicate that the proposed approach exceeds other state-of-the-art techniques in extracting weak fault features of rolling bearings.

Short Text Topic Learning Using Heterogeneous Information Network

With the explosive growth of short texts on users' preferences, learning discriminative and coherent latent topics from short texts is a critical work, since many practical applications require semantic understandings that short texts convey explicitly and implicitly. However, existing short text topic learning methods face the challenge of fully capturing semantically related co-occurrence phrases. Therefore, this paper proposes a novel Heterogeneous Information Network-based Short Text Topic learning approach (HIN-ShoTT) in terms of parts of speech, without depending on any auxiliary information. Specifically, HIN-ShoTT can be decomposed into three phases: i) seeking semantic relations among words, where HIN-ShoTT models multiple semantic relations among words based on a Heterogeneous Information Network (HIN) in terms of parts of speech; ii) extracting co-occurrence phrases and filtering noises, where HIN-ShoTT defines parts-of-speech meta structures to guide co-occurrence phrase extraction and a self-adapting threshold filtering module is proposed for discarding noises; and iii) inferring topics, where HIN-ShoTT models the generative process of co-occurrence phrases to make topic learning effective with the abundant corpus-level information. Our experimental results on three real-world datasets not only show that HIN-ShoTT performs well, but also demonstrate that it is feasible to incorporate HIN into short text topic learning for accuracy improvement.

A Hilbert–Huang Transform-Based Adaptive Fault Detection and Classification Method for Microgrids

Fault detection in microgrids is of great significance for power systems’ safety and stability. Due to the high penetration of distributed generations, fault characteristics become different from those of traditional fault detection. Thus, we propose a new fault detection and classification method for microgrids. Only current information is needed for the method. Hilbert–Huang Transform and sliding window strategy are used in fault characteristic extraction. The instantaneous phase difference of current high-frequency component is obtained as the fault characteristic. A self-adaptive threshold is set to increase the detection sensitivity. A fault can be detected by comparing the fault characteristic and the threshold. Furthermore, the fault type is identified by the utilization of zero-sequence current. Simulations for both section and system have been completed. The instantaneous phase difference of the current high-frequency component is an effective fault characteristic for detecting ten kinds of faults. Using the proposed method, the maximum fault detection time is 13.8 ms and the maximum fault type identification time is 14.8 ms. No misjudgement happens under non-fault disturbance conditions. The simulations indicate that the proposed method can achieve fault detection and classification rapidly, accurately, and reliably.

High impedance fault detection method for distribution networks under non-linear conditions

This paper presents a new high impedance fault (HIF) detection method considering the presence of non-linear loads in the distribution network. It adopts a cumulative error-based index to monitor the third harmonic phase angle, which is extracted with Stockwell transform and compared to a self-adaptive threshold for fault detection and classification. The proposed method requires only a single measurement device on the substation. Results show that the proposed method is able to quickly detect HIF in all simulated cases, considering several fault locations and different contact surfaces, not being affected by conditions such as fault inception angle, system loading percentage, or fault location. Additionally, results suggest that it can distinguish HIF from capacitor banks switching, low impedance faults, transformer energizing, and feeder energizing along with linear and non-linear loads switching. HIF real oscillographic records are also used for validation. The proposed method achieved a fast detection, in less than 11 cycles for most cases, and a high detection rate, over 99%, with no false positives.