Adaptive Window Research Articles

Enhancement based convolutional dictionary network with adaptive window for low-dose CT denoising.

Recently, one promising approach to suppress noise/artifacts in low-dose CT (LDCT) images is the CNN-based approach, which learns the mapping function from LDCT to normal-dose CT (NDCT). However, most CNN-based methods are purely data-driven, thus lacking sufficient interpretability and often losing details. To solve this problem, we propose a deep convolutional dictionary learning method for LDCT denoising, in which a novel convolutional dictionary learning model with adaptive window (CDL-AW) is designed, and a corresponding enhancement-based convolutional dictionary learning network (called ECDAW-Net) is constructed to unfold the CDL-AW model iteratively using the proximal gradient descent technique. In detail, the adaptive window-constrained convolutional dictionary atom is proposed to alleviate spectrum leakage caused by data truncation during convolution. Furthermore, in the ECDAW-Net, a multi-scale edge extraction module that consists of LoG and Sobel convolution layers is proposed in the unfolding iteration, to supplement lost textures and details. Additionally, to further improve the detail retention ability, the ECDAW-Net is trained by the compound loss function of the pixel-level MSE loss and the proposed patch-level loss, which can assist to retain richer structural information. Applying ECDAW-Net to the Mayo dataset, we obtained the highest peak signal-to-noise ratio (33.94) and sub-optimal structural similarity (0.92). Compared with some state-of-art methods, the interpretable ECDAW-Net performs well in suppressing noise/artifacts and preserving textures of tissue.

Abnormal Data Detection Based on Adaptive Sliding Window and Weighted Multiscale Local Outlier Factor for Machinery Health Monitoring

Identifying abnormal data to improve data quality is of great importance for machinery health monitoring (MHM). Existing abnormal data detection methods generally depend on appropriate parameter settings and prior knowledge of data distribution, which result in relatively low adaptability to MHM data. To obtain more reliable MHM results, this paper proposed a novel method to detect abnormal data. First, the concept of adaptive sliding window (ASW) is defined and the advantages of ASW in avoiding data leakage and data redundancy are derived. The ASW can optimally divide the overall data into several segments. Next, statistical factors in time- and frequency-domain are extracted from these segments to generate corresponding research objects. Then, an improved weighted multiscale local outlier factor (WMLOF) algorithm is proposed herein to evaluate the data anomaly degree of each object. The WMLOF has the ability to assess and fuse the local outlier factor characteristics at multiple scales, and eventually yields a more comprehensive WMLOF value to evaluate the anomaly degree of the MHM data. Finally, the effectiveness and superiority of the ASW and WMLOF are validated by a synthetic simulation of a faulty rolling bearing and two engineering projects pertaining to a railway vehicle gearbox, and bench test data. Comparative analysis shows that the proposed abnormal data detection method based on the ASW and WMLOF strategies outperforms five reported algorithms in outlier detection.

Automated quantification of small defects in ultrasonic phased array imaging using AWGA-gcForest algorithm

ABSTRACT The imaging quality of ultrasonic phased array technology significantly affects the quantification of small defects inside metals. Moreover, traditional manual evaluation methods require a substantial amount of manpower and time, and they are easily influenced by subjective factors. To mitigate the impact of imaging quality on the automated quantification of defects, this study proposes an improved gcForest algorithm (AWGA-gcForest, Adaptive window genetic algorithm- gcForest). Firstly, the acquired ultrasonic full matrix data from the phased array is processed using full focusing A-scan technique. Additionally, a window averaging weighting method is introduced for ultrasound signal processing to eliminate the interference from incident waves and backwall echoes. Subsequently, a multi-scale adaptive sliding window adjustment strategy based on Mean Absolute Deviation (MAD) is employed to reduce the computational time complexity of the gcForest algorithm. Furthermore, a weighted approach optimised with a genetic algorithm is utilised in the cascade forest to improve the classification accuracy and generalisation capability. Experimental results demonstrate that the proposed method achieves an accuracy of 97.50%, precision of 97.26%, recall rate of 96.63%, and F1 score of 96.92, exhibiting higher detection accuracy compared to other models while also reducing time costs.

High-resolution sampling aliasing frequency distribution

Rotating blades are core but fragile working parts in turbomachinery, and damage to the rotating blades severely endangers operational safety; thus, it is critical to monitor the blade condition to find the incipient failure. Blade tip timing (BTT) is a potential measurement for rotating blades owing to its non-contact nature, high efficiency, and long service life. Identifying vibration parameters for condition monitoring is a core issue. Most existing BTT methods rely on multi-probe layouts to overcome undersampling to extract parameters. However, restricted by weight, safety, installation, and maintenance costs, it is desired to extract vibration parameters on the premise of the minimal layout. Motivated by this demand, we propose a single probe-based parameter identification method, where sampling aliasing frequency (SAFE) distribution plays a significant role. To improve the accuracy of parameter identification, a synchrosqueezing adaptive window length short-time Fourier transform (SS-AWSTFT) is proposed to construct a high-resolution SAFE distribution. Moreover, a new fitting model considering rotational effect is developed to extract engine order (EO), static frequency, and dynamic frequency coefficient from the SFAE distribution. Owing to considering the rotational effect, the new fitting model performs better in parameter estimation. Finally, the effectiveness and robustness of the proposed method were demonstrated by simulations and experiments.

Adaptive dynamic windowing approach based on risk degree function

Aiming at the problem that the dynamic window approach (DWA) has low planning efficiency and it is difficult to avoid fast-moving obstacles, an improved adaptive DWA based on risk degree function is proposed in this paper. The risk function is designed and introduced into the evaluation function of the traditional DWA to evaluate the risk of collision between dynamic obstacles and the robot, so that the robot can effectively avoid faster obstacles. Then, according to the fuzzy control principle, the adaptive weight coefficient is designed to improve the evaluation function, so that the mobile robot can move to the target point more efficiently. The simulation results show that compared with the traditional DWA, the adaptive DWA based on risk degree function reduces the time of completing the planning task by about 8%, and the path length after the completion of the planning by about 8%, it indicates that the improved algorithm has higher efficiency. After 50 repeated experiments, using the adaptive DWA based on risk degree function successfully avoids all obstacles to complete the planning task, which shows that this algorithm has higher security.

A high-precision online compensation method for random errors of optical gyroscope

The traditional Allan variance cannot reflect the time-varying characteristics of random error signals, and the dynamic Allan variance cannot simultaneously guarantee both the signal dynamic tracking performance and confidence of variance estimation due to adopting the sliding window with fixed window length to intercept and segment random error signals. To address these issues, a dynamic Allan variance analysis method with adaptive window length is proposed. Based on the idea of piecewise stability of time-varying random error signals, a sliding window is introduced to intercept and segment the signal to be analyzed along the time axis, and the Allan variance analysis results of all signal segments are synthesized to track and describe the time-varying characteristics of the signal. Then according to the changes in the statistical characteristics of the random error signal, an adaptive factor is constructed to adaptively adjust the length of the sliding window, thereby guaranteeing both the signal dynamic tracking performance and confidence of variance estimation at the same time. On this basis, this method is applied to the online compensation of optical gyroscope random errors, and then a high-precision online compensation method for random errors of optical gyroscope is proposed. The experimental results show that the proposed method can achieve high-precision online compensation for the time-varying random error signals of optical gyroscope measured in complex environment such as uncertainty and noise interference.

HRSHO: A hybrid rain optimized spotted hyena optimizer for efficient feature selection in CNN-based sinusitis classification

The optimal classification of sinusitis is a critical need in medical field due to prevalence and impact of sinusitis on individuals' health. Sinusitis diagnosis has relied on clinical examination, patient history and imaging techniques such as computed tomography scans or magnetic resonance imaging. With advancements in artificial intelligence and machine learning, there has been growing interest in developing automated and data-driven approaches for sinusitis classification. The proposed Deep Learning (DL) based detection model uses Adaptive Gaussian Notch Filter (AGNF) for pre-processing the input CT scans of paranasal sinus. Using adaptive window size variation, AGNF defuses noisy peaks in input image. With the application of Recurrent Neural Network (RNN), relevant Region of Interests (ROIs) that buttress the accurate sinusitis classification are segmented separately. The critical features including contrast, ASM, energy, dissimilarity and homogeneity are extracted from segmented image using Gray-Level Co-occurrence Matrix (GLCM). A novel hybrid feature selection algorithm combining Spotted Hyena Optimization (SHO) and Rain Optimization Algorithm (ROA) is introduced in this work for supporting optimal classification of sinusitis. The proposed Hybrid Rain Optimized Spotted Hyena Optimizer (HRSHO) seeks to identify the most salient and discriminative features among extracted set, thereby reducing feature dimensionality and improving classification performance. The proposed DL based sinusitis is examined and executed in Python software. Moreover, from the derived results it is noted that proposed HRSHO operates as an excellent feature selection technique, as it selects minimum number of features by excluding redundant features with significantly higher accuracy and minimum computational cost.

Defocus compensation scheme for femtosecond laser direct writing of large-area non-periodic micro-structures

The optical path error, the position-independent geometric error, and the form and position tolerance of the substrate will lead to position deviations of the laser focus resulting in the inconsistency of the large area micro-structure and limiting the processing range of the femtosecond laser fabrication platform. To realize continuous defocus compensation along the processing trajectory, an autofocus scheme consisting of an optimized depth from focus (DFF) method based on grid division and the bilinear interpolation algorithm is proposed in this paper. The optimized DFF method includes the Laplace-DWT evaluation function combining the advantages of spatial domain and frequency domain sharpness evaluation function, an adaptive focusing window selection method, and an adaptive step length three-step hill-climbing search method. The optimized DFF method was used to obtain the best focus position of each grid point. Compared with the traditional DFF method, the optimized algorithm has higher focusing precision and avoids the disadvantage of time-consuming focusing. Then, to compensate for interpolating points of the processing trajectory, the bilinear interpolation algorithm was introduced to achieve online focusing during two-photon lithography processing. Finally, the feasibility and effectiveness of the auto-focusing scheme were validated by experiments. Through defocus compensation, the sharpness evaluation value of the focus image on the processing trajectory improved by 90%, which indicated the successful focus of the laser beam on the workpiece surface. Moreover, a high-precision and large-area encoded micro-polarizer array with consistent structure were fabricated successfully. This paper provides a simple way to identify and compensate the defocus errors online with high accuracy and efficiency. The proposed method can improve the machining accuracy and quality of the femtosecond laser direct writing.

A noise reduction method of rolling bearing based on empirical wavelet transform and adaptive time frequency peak filtering

The vibration signal of rolling bearing with variable operating conditions contains complex interference components, which will cause low fault diagnosis accuracy, especially in strong noise case. To solve this problem, we proposed a noise reduction method of rolling bearing with variable operating based on empirical wavelet transform and adaptive time-frequency peak filtering (EWT-ATFPF). Firstly, empirical wavelet transform is used to obtain different frequency intrinsic mode functions (IMFs). Secondly, a modified adaptive window length formula for time-frequency peak filtering (TFPF) is constructed by combining the sampling ratio index and a fault sensitivity indicator that calculated by kurtosis and correlation coefficients of IMFs, which can better characterize the impact components. Thirdly, to balance noise reduction effect and the fidelity of IMFs, we proposed an improved TFPF method by adaptively adjusting its windows length. The adaptive method could be carried out using the proposed fault sensitivity indicator and window length formula, and the denoising IMFs could be obtained by ATFPF. Finally, the denoising vibration signal is reconstructed by using the denoising IMFs. The performance of fault diagnosis of the proposed method is verified by using simulated signal and bearing fault test data. The results show that the proposed EWT-ATFPF method could effectively achieve noise reduction under variable operating conditions.

A novel Edge architecture and solution for detecting concept drift in smart environments

The proliferation of the Internet of Things (IoT), artificial intelligence (AI), the adoption of 5G, and progress towards 6G technology have led to the accumulation of massive amounts of real-world data; however, a significant portion of the data generated by smart cities and smart buildings remains unused. A notable problem is the shift of statistical properties in real-world streaming over time caused by unexpected factors, referred to as concept drift, which results in less efficient predictive models. To address this problem, the latest research leverages the cloud–edge continuum paradigm for the deployment of AI and general smart city applications while utilising the available resources optimally. In this article, we propose a computing architecture for different smart city applications in edge micro data centre (EMDC) settings over a hybrid cloud–edge continuum to support the deployment of AI workloads. We implement a feedback-driven automated concept drift detection and adaptation methodology, combining base learner long short-term memory (LSTM) with Page–Hinkley test (PHT), adaptive windowing (ADWIN) and the Kolmogorov–Smirnov windowing (KSWIN). Real-world data streams are utilised to forecast from various environmental sensors installed at the University of Oulu Smart Campus. The feedback-based concept drift detection and adaption process is first evaluated using synthetic datasets with known concept drift points and then employed in the real-world data. Subsequently, the implementation is evaluated using the state-of-the-art MAE, RMSE, and MAPE methods. The results showed a reduction in MAPE from 8.5% to 3.88% when concept drift detection was applied. Additionally, the challenges faced and the effectiveness of the suggested solutions are explored.

Spatial graph attention network-based object tracking with adaptive cosine window

Spatial graph attention network-based object tracking with adaptive cosine window

Adaptive region aggregation for multi‐view stereo matching using deformable convolutional networks

AbstractDeep‐learning methods have demonstrated promising performance in multi‐view stereo (MVS) applications. However, it remains challenging to apply a geometrical prior on the adaptive matching windows to achieve efficient three‐dimensional reconstruction. To address this problem, this paper proposes a learnable adaptive region aggregation method based on deformable convolutional networks (DCNs), which is integrated into the feature extraction workflow for MVSNet method that uses coarse‐to‐fine structure. Following the conventional pipeline of MVSNet, a DCN is used to densely estimate and apply transformations in our feature extractor, which is composed of a deformable feature pyramid network (DFPN). Furthermore, we introduce a dedicated offset regulariser to promote the convergence of the learnable offsets of the DCN. The effectiveness of the proposed DFPN is validated through quantitative and qualitative evaluations on the BlendedMVS and Tanks and Temples benchmark datasets within a cross‐dataset evaluation setting.

DAWTran: dynamic adaptive windowing transformer network for pneumothorax segmentation with implicit feature alignment

Objective. This study aims to address the significant challenges posed by pneumothorax segmentation in computed tomography images due to the resemblance between pneumothorax regions and gas-containing structures such as the trachea and bronchus. Approach. We introduce a novel dynamic adaptive windowing transformer (DAWTran) network incorporating implicit feature alignment for precise pneumothorax segmentation. The DAWTran network consists of an encoder module, which employs a DAWTran, and a decoder module. We have proposed a unique dynamic adaptive windowing strategy that enables multi-head self-attention to effectively capture multi-scale information. The decoder module incorporates an implicit feature alignment function to minimize information deviation. Moreover, we utilize a hybrid loss function to address the imbalance between positive and negative samples. Main results. Our experimental results demonstrate that the DAWTran network significantly improves the segmentation performance. Specifically, it achieves a higher dice similarity coefficient (DSC) of 91.35% (a larger DSC value implies better performance), showing an increase of 2.21% compared to the TransUNet method. Meanwhile, it significantly reduces the Hausdorff distance (HD) to 8.06 mm (a smaller HD value implies better performance), reflecting a reduction of 29.92% in comparison to the TransUNet method. Incorporating the dynamic adaptive windowing (DAW) mechanism has proven to enhance DAWTran’s performance, leading to a 4.53% increase in DSC and a 15.85% reduction in HD as compared to SwinUnet. The application of the implicit feature alignment (IFA) further improves the segmentation accuracy, increasing the DSC by an additional 0.11% and reducing the HD by another 10.01% compared to the model only employing DAW. Significance. These results highlight the potential of the DAWTran network for accurate pneumothorax segmentation in clinical applications, suggesting that it could be an invaluable tool in improving the precision and effectiveness of diagnosis and treatment in related healthcare scenarios. The improved segmentation performance with the inclusion of DAW and IFA validates the effectiveness of our proposed model and its components.

A novel non-intrusive load monitoring method based on ResNet-seq2seq networks for energy disaggregation of distributed energy resources integrated with residential houses

The increasing effects of global warming and energy depletion have raised concerns about the pollution caused by traditional oil and fossil energy usage. Distributed energy resources (DERs) have emerged as promising techniques to address these issues. However, the growing proportion of power injection from DERs presents technical challenges such as power quality decline and power flow reversal, which may destabilize the grid. Non-intrusive load monitoring (NILM) is a promising and cost-effective approach that can provide residential power information to improve grid scheduling, dispatching, and optimize residential power consumption behavior. This paper proposes a novel NILM method that can decompose the power of both household appliances and DERs integrated with residential houses. The proposed approach employs a data segment method that utilizes adaptive window lengths to identify the behavior of DERs based on the operating periods and characteristics of different equipment. A non-intrusive load monitoring method is proposed based on the ResNet-seq2seq framework to address the problem of model degradation and gradient vanishing. Post-processing techniques are applied to the output results using a feature list of the target equipment. Experiments are conducted on a dataset organized based on the REDD and Pecan Street datasets to demonstrate the accuracy and effectiveness of the proposed method. The results show that compared to the start-of-the-art CNN-seq2seq model, the proposed method achieved 54.95%, 69.18%, 56.95%, and 84.98% reduction in MAE for PV, EV, refrigerators, and microwaves, respectively.

Research on the Improvement of Semi-Global Matching Algorithm for Binocular Vision Based on Lunar Surface Environment.

The low light conditions, abundant dust, and rocky terrain on the lunar surface pose challenges for scientific research. To effectively perceive the surrounding environment, lunar rovers are equipped with binocular cameras. In this paper, with the aim of accurately detect obstacles on the lunar surface under complex conditions, an Improved Semi-Global Matching (I-SGM) algorithm for the binocular cameras is proposed. The proposed method first carries out a cost calculation based on the improved Census transform and an adaptive window based on a connected component. Then, cost aggregation is performed using cross-based cost aggregation in the AD-Census algorithm and the initial disparity of the image is calculated via the Winner-Takes-All (WTA) strategy. Finally, disparity optimization is performed using left-right consistency detection and disparity padding. Utilizing standard test image pairs provided by the Middleburry website, the results of the test reveal that the algorithm can effectively improve the matching accuracy of the SGM algorithm, while reducing the running time of the program and enhancing noise immunity. Furthermore, when applying the I-SGM algorithm to the simulated lunar environment, the results show that the I-SGM algorithm is applicable in dim conditions on the lunar surface and can better help a lunar rover to detect obstacles during its travel.

Modified Strong Tracking Slide Window Variational Adaptive Kalman Filter With Unknown Noise Statistics

The filter performance will be degraded in the measurements with time-varying and unknown noise statistics. To combat the above challenges, a modified strong tracking slide window variational adaptive Kalman filter algorithm is proposed in this article. Firstly, the multiple fading factors are integrated into the proposed algorithm to adjust the error covariance. Next, an improved adaptive slide window method is designed for variational Bayesian Kalman filtering by adaptively adjusting the slide window size and correcting the previous state according to the later state, which improves the estimation accuracy and computational efficiency. Finally, the inverse Wishart distribution is considered for modeling process and measurement noise, and the state vector, as well as noise statistics, are inferred via the variational Bayesian technique without prior noise covariance information. Simulation results demonstrate that the proposed filter algorithm is more robust than existing filters in counteracting measurement and process noise uncertainties.

Influence of Image Processing Method on Wavefront Reconstruction Accuracy of Large-Aperture Laser

In order to improve the wavefront reconstruction accuracy of a large-aperture laser, this paper proposed an adaptive window preprocessing algorithm based on the threshold center of gravity method (AW-TCoG). The effects of median filtering and mean filtering on spot image processing and wavefront reconstruction accuracy are simulated and analyzed. The results show that the mean filtering method has a better effect on noise elimination and can further improve the accuracy of wavefront reconstruction. In addition, the centroid detection errors of large-aperture laser wavefront reconstruction through the center of gravity (CoG), the threshold center of gravity (T-CoG), and the Windowing method were studied. The analysis shows that, due to the influence of noise, the wavefront reconstruction accuracy is poor when the CoG and Windowing methods are used to calculate centroid parameters, while the wavefront reconstruction accuracy of the threshold centroid method is better and can reach 0.2λ. When using the AW-TCoG proposed in this paper, the wavefront reconstruction accuracy can be maintained within 0.1λ for different incident wavefront RMS values and spot images with different signal-to-noise ratio (SNR) levels. Compared with the traditional threshold centroid method, the wavefront reconstruction accuracy of this method is significantly improved.

An improved local maximum synchrosqueezing transform with adaptive window width for instantaneous frequency identification of time-varying structures

Civil engineering structures usually exhibit time-varying characteristics under operational conditions due to changes in environment and loads. The identification of instantaneous frequencies (IFs) from measured responses of time-varying structures is a challenge when using the local maximum synchrosqueezing transform (LMSST) due to the difficulty in selecting the appropriate window width. In this study, an improved LMSST with adaptive window width (ALMSST) is proposed to circumvent the limitations of LMSST. The window width of ALMSST can be adaptively determined by combining the autoregressive power spectrum-based variational modal decomposition (AR-VMD) and a window width optimization algorithm. The AR-VMD is used to decompose the multi-component signal into mono-component signals. The Rényi entropy is adopted as an evaluation index in the window width optimization algorithm for selecting the optimal window width for each mono-component signal. Therefore, ALMSST can provide a highly concentrated time–frequency (TF) representation for all mono-component signals. Two simulated signals demonstrate that the ALMSST improves the accuracy of identified IFs compared with LMSST. Numerical simulation of a three-story shear building model with time-varying stiffness shows that ALMSST can accurately identify the IFs of time-varying structures under heavy noise. A cable test with linear and sinusoidal varying tension forces and a vehicle-bridge interaction (VBI) model system with different vehicle weights are investigated to verify the applicability of ALMSST to track the IFs of time-varying structures. Numerical simulation and experimental results illustrate that ALMSST performs well in identifying the IFs of time-varying structures.

Ensemble deep clustering analysis for time window determination of event-related potentials

ObjectiveCluster analysis of spatio-temporal event-related potential (ERP) data is a promising tool for exploring the measurement time window of ERPs. However, even after preprocessing, the remaining noise can result in uncertain cluster maps followed by unreliable time windows while clustering via conventional clustering methods. MethodsWe designed an ensemble deep clustering pipeline to determine a reliable time window for the ERP of interest from temporal concatenated grand average ERP data. The proposed pipeline includes semi-supervised deep clustering methods initialized by consensus clustering and unsupervised deep clustering methods with end-to-end architectures. Ensemble clustering from those deep clusterings was used by the designed adaptive time window determination to estimate the time window. ResultsAfter applying simulated and real ERP data, our method successfully obtained the time window for identifying the P3 components (as the interest of both ERP studies) while additional noise (e.g., adding 20 dB to −5 dB white Gaussian noise) was added to the prepared data. ConclusionCompared to the state-of-the-art clustering methods, a superior clustering performance was yielded from both ERP data. Furthermore, more stable and precise time windows were elicited as the noise increased. SignificanceOur study provides a complementary understanding of identifying the cognitive process using deep clustering analysis to the existing studies. Our finding suggests that deep clustering can be used to identify the ERP of interest when the data is imperfect after preprocessing.

BATFormer: Towards Boundary-Aware Lightweight Transformer for Efficient Medical Image Segmentation.

Transformers, born to remedy the inadequate receptive fields of CNNs, have drawn explosive attention recently. However, the daunting computational complexity of global representation learning, together with rigid window partitioning, hinders their deployment in medical image segmentation. This work aims to address the above two issues in transformers for better medical image segmentation. We propose a boundary-aware lightweight transformer (BATFormer) that can build cross-scale global interaction with lower computational complexity and generate windows flexibly under the guidance of entropy. Specifically, to fully explore the benefits of transformers in long-range dependency establishment, a cross-scale global transformer (CGT) module is introduced to jointly utilize multiple small-scale feature maps for richer global features with lower computational complexity. Given the importance of shape modeling in medical image segmentation, a boundary-aware local transformer (BLT) module is constructed. Different from rigid window partitioning in vanilla transformers which would produce boundary distortion, BLT adopts an adaptive window partitioning scheme under the guidance of entropy for both computational complexity reduction and shape preservation. BATFormer achieves the best performance in Dice of 92.84 %, 91.97 %, 90.26 %, and 96.30 % for the average, right ventricle, myocardium, and left ventricle respectively on the ACDC dataset and the best performance in Dice, IoU, and ACC of 90.76 %, 84.64 %, and 96.76 % respectively on the ISIC 2018 dataset. More importantly, BATFormer requires the least amount of model parameters and the lowest computational complexity compared to the state-of-the-art approaches. Our results demonstrate the necessity of developing customized transformers for efficient and better medical image segmentation. We believe the design of BATFormer is inspiring and extendable to other applications/frameworks.