Representative Feature Set Research Articles

Weed detection and recognition in complex wheat fields based on an improved YOLOv7.

The precise detection of weeds in the field is the premise of implementing weed management. However, the similar color, morphology, and occlusion between wheat and weeds pose a challenge to the detection of weeds. In this study, a CSCW-YOLOv7 based on an improved YOLOv7 architecture was proposed to identify five types of weeds in complex wheat fields. First, a dataset was constructed for five weeds that are commonly found, namely, Descurainia sophia, thistle, golden saxifrage, shepherd's purse herb, and Artemisia argyi. Second, a wheat weed detection model called CSCW-YOLOv7 was proposed to achieve the accurate identification and classification of wheat weeds. In the CSCW-YOLOv7, the CARAFE operator was introduced as an up-sampling algorithm to improve the recognition of small targets. Then, the Squeeze-and-Excitation (SE) network was added to the Extended Latent Attention Networks (ELAN) module in the backbone network and the concatenation layer in the feature fusion module to enhance important weed features and suppress irrelevant features. In addition, the contextual transformer (CoT) module, a transformer-based architectural design, was used to capture global information and enhance self-attention by mining contextual information between neighboring keys. Finally, the Wise Intersection over Union (WIoU) loss function introducing a dynamic nonmonotonic focusing mechanism was employed to better predict the bounding boxes of the occluded weed. The ablation experiment results showed that the CSCW-YOLOv7 achieved the best performance among the other models. The accuracy, recall, and mean average precision (mAP) values of the CSCW-YOLOv7 were 97.7%, 98%, and 94.4%, respectively. Compared with the baseline YOLOv7, the improved CSCW-YOLOv7 obtained precision, recall, and mAP increases of 1.8%, 1%, and 2.1%, respectively. Meanwhile, the parameters were compressed by 10.7% with a 3.8-MB reduction, resulting in a 10% decrease in floating-point operations per second (FLOPs). The Gradient-weighted Class Activation Mapping (Grad-CAM) visualization method suggested that the CSCW-YOLOv7 can learn a more representative set of features that can help better locate the weeds of different scales in complex field environments. In addition, the performance of the CSCW-YOLOv7 was compared to the widely used deep learning models, and results indicated that the CSCW-YOLOv7 exhibits a better ability to distinguish the overlapped weeds and small-scale weeds. The overall results suggest that the CSCW-YOLOv7 is a promising tool for the detection of weeds and has great potential for field applications.

Effective Video Summarization Using Channel Attention-Assisted Encoder–Decoder Framework

A significant number of cameras regularly generate massive amounts of data, demanding hardware, time, and labor resources to acquire, process, and monitor. Asymmetric frames within videos pose a challenge to automatic summarization of videos, making it challenging to capture key content. Developments in computer vision have accelerated the seamless capture and analysis of high-resolution video content. Video summarization (VS) has garnered considerable interest due to its ability to provide concise summaries of lengthy videos. The current literature mainly relies on a reduced set of representative features implemented using shallow sequential networks. Therefore, this work utilizes an optimal feature-assisted visual intelligence framework for representative feature selection and summarization. Initially, the empirical analysis of several features is performed, and ultimately, we adopt a fine-tuning InceptionV3 backbone for feature extraction, deviating from conventional approaches. Secondly, our strategic encoder–decoder module captures complex relationships with five convolutional blocks and two convolution transpose blocks. Thirdly, we introduced a channel attention mechanism, illuminating interrelations between channels and prioritizing essential patterns to grasp complex refinement features for final summary generation. Additionally, comprehensive experiments and ablation studies validate our framework's exceptional performance, consistently surpassing state-of-the-art networks on two benchmarks (TVSum and SumMe) datasets.

A family of recognition algorithms based on the construction of k-dimensional threshold rules

In this article, the authors proposed a family of recognition algorithms based on k-dimensional threshold rules and focused on the object recognition presented in the feature space of high dimension. These recognition algorithms are based on the concept of forming a set of k-dimensional preferred threshold rules that define an extremal algorithm within the proposed family of algorithms. A structural description of the developed family of recognition algorithms is given, presented as a sequence of computational procedures. The main ones are the procedures for determining a set of representative features; discriminant functions for each class; normalizing transformations; groups of highly correlated discriminant functions; basic discriminant functions; a set of preferred discriminant functions; a generalized recognition operator based on preferred discriminant functions. The results of experimental studies are presented; they showed that an increase in the accuracy of recognition results is provided (in comparison with known analogs) when using the indicated family of recognition algorithms. The main advantage of the proposed recognition algorithms is not only an increase in accuracy but also a significant reduction in arithmetic operations in object recognition. This property of the algorithms makes it possible to create various software recognition systems that operate in real time.

Realizing accurate battery capacity estimation using 4 min 1C discharging data

Accurate capacity estimation is important to ensure the safe operation of battery. Current data-driven capacity estimation methods mainly rely on large volume charging or discharging data, which usually lasts for tens of minutes to hours, to extract effective battery aging features. The large volume data requirement restricts the application in real-world scenarios to some extent. In this paper, a novel capacity estimation method, which only requires several-minute discharging data to realize accurate capacity estimation, is proposed. Firstly, the idea of regression shapelet is introduced and shapelet distance is defined to capture the battery degradation trend. Then maximum relevance minimum redundancy algorithm is used to select the representative shapelet feature set, which not only has strong correlation with battery capacity, but also contains least redundancy among each other. Finally, eXtreme gradient boosting is adopted as the final regressor mapping from shapelet distance to battery capacity. The proposed method is verified on the public Oxford battery dataset. Results show that 4min 1C discharging data is sufficient for the proposed method to realize accurate capacity estimation with mean absolute relative error of 0.86%, which is superior to existing methods in terms of data volume requirement or accuracy. Moreover, the proposed method is promising to be applied in vehicle-to-grid scenarios considering its light computational burden.

In-field rice panicles detection and growth stages recognition based on RiceRes2Net

Accurate rice panicle detection and growth stages recognition are crucial steps in rice field phenotyping. However, conventional manual characterization of rice panicles is time consuming and labor intensive. In this study, a RiceRes2Net based on improved Cascade RCNN (Region-CNN) architecture was proposed to detect the rice panicle and recognize the growth stages under the complex field environment. RiceRes2Net first adopted the Res2Net network and Feature Pyramid Network （FPN） as the backbone network to generate and fuse multi-scale feature maps. Then, RiceRes2Net constituted a four IoU thresholds cascade RCNN to deal with multi-scale feature maps to give the target class prediction and coordinate regression of the bounding boxes. In addition, Soft non-maximum suppression (Soft NMS) and Generalized Intersection over Union (GIoU) loss function were also integrated into RiceRes2Net to better predict the bounding boxes of the occluded panicles. Datasets of the rice panicles were acquired by smartphone in two comprehensive field plot experiments under complex field background. Rice panicles differed in genotype, planting density, growing practices, planting season and growth stages, which constituted a comprehensive rice panicles phenotyping. The results showed that RiceRes2Net outperformed the traditional cascade RCNN in rice panicle detection, with average precision (AP) values of 96.8%, 93.7%, 82.4% at booting stage, heading stage, and filling stage, respectively. Furthermore, RiceRes2Net has a significant advantage in detecting the occlusion panicle thereby increase the accuracy. To test the robustness of RiceRes2Net, the counting results of RiceRes2Net was compared with the manual counting results with an independent test set. The RMSE values at three growth stages were 1.19, 2.56, and 3.13, respectively. In addition, the performance of the RiceRes2Net was compared to the widely used state-of-art deep learning models. The results showed that RiceRes2Net can learn a more representative set of features that helped better locate the rice panicles at three growth stages, and thus achieved better detection accuracy than the other deep learning models. In terms of panicle growth stages recognition, RiceRes2Net showed satisfactory results with high precision values of 99.83%, 99.34%, and 94.59% in recognition of booting stage, heading stage, and filling stage, respectively. The average accuracy of growth stages recognition was 96.42%. The overall results suggest that RiceRes2Net is a promising tool for detection of rice panicles and the growth stage, and has great potentials for field applications.

Deep Encrypted Traffic Detection: An Anomaly Detection Framework for Encryption Traffic Based on Parallel Automatic Feature Extraction.

With an increasing number of network attacks using encrypted communication, the anomaly detection of encryption traffic is of great importance to ensure reliable network operation. However, the existing feature extraction methods for encrypted traffic anomaly detection have difficulties in extracting features, resulting in their low efficiency. In this paper, we propose a framework of encrypted traffic anomaly detection based on parallel automatic feature extraction, called deep encrypted traffic detection (DETD). The proposed DETD uses a parallel small-scale multilayer stack autoencoder to extract local traffic features from encrypted traffic and then adopts an L1 regularization-based feature selection algorithm to select the most representative feature set for the final encrypted traffic anomaly detection task. The experimental results show that DETD has promising robustness in feature extraction, i.e., the feature extraction efficiency of DETD is 66% higher than that of the conventional stacked autoencoder, and the anomaly detection performance is as high as 99.998%, and thus DETD outperforms the deep full-range framework and other neural network anomaly detection algorithms.

Unsupervised Ship Detection in SAR Images Using Superpixels and CSPNet

Ship detection in synthetic aperture radar (SAR) images is critical to the ocean surveillance and rescue. Although many deep learning SAR ship detection methods have been proposed, the performance of these methods depends on the size and quality of the training samples. To resolve these issues, this letter presents an unsupervised ship detection method in SAR images using superpixel segmentation and cross stage partial network (CSPNet). First, the SAR image is over-segmented into superpixels based on our previously proposed superpixel generation algorithm. Then, the complex signal kurtosis (CSK) and a local superpixel contrast are integrated as a statistical indicator for the automatic identification of ship superpixels and background superpixels, thus leading to generation of training samples. Finally, the segmented superpixels are input to the CSPNet, which can learn a representative feature set with high discrimination ability between ships and backgrounds. Our method can achieve pixel-level detection map rather than the bounding box result. Experiments based on the Gaofen-3 and TerraSAR SAR data demonstrate that our method can achieve above 90% actual detection rate.

An Automated Multi-Tab Website Fingerprinting Attack

In Website Fingerprinting (WF) attack, a local passive eavesdropper utilizes network flow information to identify which web pages a user is browsing. Previous researchers have demonstrated the feasibility and effectiveness of WF attacks under a strong Single Page Assumption: the network flow extracted by the adversary belongs to a single web page. In reality, the assumption may not hold because users tend to open multiple tabs simultaneously (or within a short period of time) so that their network traffic is mixed. In this article, we propose an automated multi-tab Website Fingerprinting attack that is able to accurately classify websites regardless of the number of simultaneously opened pages. Our design is powered by two innovative designs. First, we develop a split point classification method to dynamically identify the split point between the first page and its subsequent pages. As a result, the network traffic before the split point is solely generated for the first page. Then, we propose a new chunk-based WF classifier to infer the websites based on the initial chunk of clean traffic. For both classifiers, we apply automated feature selection to select a concise yet representative feature set. We implement a prototype of our design and perform extensive evaluations using SSH and Tor-based datasets to demonstrate the effectiveness of both our system components individually and the integrated system as a whole.

State of Health and Lifetime Prediction of Lithium-ion Batteries Using Self-learning Incremental Models

Lithium-ion batteries are key energy storage elements in the context of environmental-aware energy systems representing a crucial technology to achieve the goal of zero carbon emission. Therefore, its conditions must be monitored to guarantee the safe and reliable operation of the systems that use these components. Furthermore, lithium-ion batteries’ prognostics and health management policies must cope with the nonlinear and time-varying nature of the complex electrochemical dynamics of battery degradation. This paper proposes an incremental-learning-based algorithm to estimate the State of Health (SoH) and the Remaining Useful Life (RUL) of lithium-ion batteries based on measurement data streams. For this purpose, a two-layer framework is proposed based on incremental modeling of the SoH. In the first layer, a set of representative features are extracted from voltage and current data of partial charging and discharging cycles; these features are then used to train the proposed model in a recursive procedure to estimate the battery’s SoH. The second layer uses the capacity data for incremental learning of an Autoregressive (AR) model for the SoH, which will be used to propagate the battery’s degradation through time to make the RUL prediction. The proposed method was applied to two datasets for experimental evaluation, one from CALCE and another from NASA. The proposed framework was able to estimate the SoH of 8 different lithium-ion cells with an average percentage error below 1.5% for all scenarios, while the lifetime model predicted the cell’s RUL with a maximum average error of 25%.

Melanoma segmentation: A framework of improved DenseNet77 and UNET convolutional neural network

AbstractMelanoma is the most fatal type of skin cancer which can cause the death of victims at the advanced stage. Extensive work has been presented by the researcher on computer vision for skin lesion localization. However, correct and effective melanoma segmentation is still a tough job because of the extensive variations found in the shape, color, and sizes of skin moles. Moreover, the presence of light and brightness variations further complicates the segmentation task. We have presented improved deep learning (DL)‐based approach, namely, the DenseNet77‐based UNET model. More clearly, we have introduced the DenseNet77 network at the encoder unit of the UNET approach to computing the more representative set of image features. The calculated keypoints are later segmented by the decoder of the UNET model. We have used two standard datasets, namely, the ISIC‐2017 and ISIC‐2018 to evaluate the performance of the proposed approach and acquired the segmentation accuracies of 99.21% and 99.51% for the ISIC‐2017 and ISIC‐2018 datasets, respectively. We have confirmed through both the quantitative and qualitative results that the proposed improved UNET approach is robust to skin lesions segmentation and can accurately recognize the moles of varying colors and sizes.

A lightweight data representation for phishing URLs detection in IoT environments

Phishing is a cyber-attack that exploits victims’ technical ignorance or naivety and commonly involves a Uniform Resources Locator (URL). Hence, it is advantageous to detect a phishing attack by analyzing URLs before accessing them. With the raising of the Internet of Things (IoT), phishing attacks are moving to this field due to the number of IoT devices and the amount of personal information they handle. Although several approaches were proposed for phishing attacks detection, the URL-based Machine Learning approaches obtain better performance results, but all of them are dependent on the feature set used. Contradictorily, only a few works on selecting the best-suited feature set for improving the phishing detection process have been published. The present research explores how to obtain a feature set that substantially enhances the phishing detection rate in IoT environments. Hence, a feature selection algorithm was adopted and extended for getting the most representative feature set. When Random Forest is used with the proposed data representation, the phishing URL attacks discovery rate is 99.57%.

A Novel Feature-Selection Method for Human Activity Recognition in Videos

Human Activity Recognition (HAR) is the process of identifying human actions in a specific environment. Recognizing human activities from video streams is a challenging task due to problems such as background noise, partial occlusion, changes in scale, orientation, lighting, and the unstable capturing process. Such multi-dimensional and none-linear process increases the complexity, making traditional solutions inefficient in terms of several performance indicators such as accuracy, time, and memory. This paper proposes a technique to select a set of representative features that can accurately recognize human activities from video streams, while minimizing the recognition time and memory. The extracted features are projected on a canvas, which keeps the synchronization property of the spatiotemporal information. The proposed technique is developed to select the features that refer only to progression of changes. The original RGB frames are preprocessed using background subtraction to extract the subject. Then the activity pattern is extracted through the proposed Growth method. Three experiments were conducted; the first experiment was a baseline to compare the classification task using the original RGB features. The second experiment relied on classifying activities using the proposed feature-selection method. Finally, the third experiment provided a sensitivity analysis that compares between the effect of both techniques on time and memory resources. The results indicated that the proposed method outperformed original RBG feature-selection method in terms of accuracy, time, and memory requirements.

OptFus: Optical Sensor Fusion for the Classification of Multisource Data: Application to Mineralogical Mapping

We propose a new fusion-based classification technique for optical multisource remote-sensing images called OptFus. OptFus is developed to merge and process optical imagery having different spatial and spectral resolutions. The spatial features are extracted using morphological filters from the RGB data containing high spatial resolution. A feature fusion technique is developed to combine all the sensor data in a subspace using a common set of representative features. Finally, the fused features are classified using a support vector machine to ensure a robust supervised spectral classification. The proposed method is designed to allocate varying weights to the data from various imaging sensors in the fusion process. OptFus is applied to two multisource optical datasets captured from geological drill-core samples. The classification accuracy demonstrates considerable improvements compared to the state-of-the-art. A MATLAB implementation of OptFus is available online: <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/BehnoodRasti/OptFus</uri> .

Bearing Fault Diagnosis in the Mixed Domain Based on Crossover-Mutation Chaotic Particle Swarm

The classification frameworks for fault diagnosis of rolling element bearings in rotating machinery are mostly based on analysis in a single time-frequency domain, where sensitive features are not completely extracted. To solve this problem, a new fault diagnosis technique is proposed in the mixed domain, based on the crossover-mutation chaotic particle swarm optimization support vector machine. Firstly, fault features are generated using techniques in the time domain, the frequency domain, and the time-frequency domain. Secondly, the weighted maximum relevance minimum redundancy (WMRMR) algorithm is adopted to reduce the dimension of the feature set and to establish the representative feature set. Thirdly, a new crossover-mutation strategy is suggested to reduce the local minima in optimization, and an optimization disturbance is added. Finally, the support vector machine is optimized using the improved chaotic particle swarm to improve fault classification diagnosis. The effectiveness of the proposed new bearing fault diagnostic technique is verified by experimental tests under different bearing conditions. Test results showed that the bearing fault classification accuracy of CMCPSO-SVM in the mixed domain was much higher than those in a single feature domain.

Optimizing Sentiment Classification for Arabic Opinion Texts

Meanwhile, products and services reviews’ provide a guide for potential customers allowing them to reach real knowledge about such products/services while making decisions. Sentiment classification is the task of analyzing opinions expressed in textual reviews automatically. The efficiency of this task is influenced by the set of representative features extracted from the reviews. Nevertheless, the value of extracted features lies as well in those that highly contribute to the classification process. Here comes the role of dimensionality reduction to eliminate the noise and reduce the feature high space while preserving required accuracies. The Arabic language and its datasets have inherent challenges. Besides, most sentiment classification studies integrating dimensionality reduction have focused on English texts, with only few studies conducted for other languages including Arabic. Massive amounts of Arabic data have been generated due to the huge population of the Arab world, and despite that, the aforementioned technical gaps are still existing for such language. This paper proposes a supervised learning approach for Arabic reviews sentiment classification. This approach utilizes optimized compact features that depend on a well representative feature set coupled with feature reduction techniques, which manages to guarantee high accuracy and time/space savings simultaneously. The employed feature set includes a triple combination of N-gram features and positive/negative N-grams counts features obtained after considering negation handling. The proposed approach examines two different linear transformation methods; principal component analysis (PCA) as an unsupervised transformation method and latent Dirichlet allocation (LDA) as a supervised transformation method. A spam detection process is executed prior to the learning for the purpose of increasing the classifier robustness. The proposed approach has been experimented with five Arabic opinion text datasets, of different domains and varying sizes (1.6 up to 94 K reviews). Experiments have been conducted for two-class (positive/negative sentiments) and three-class (positive/negative/neutral sentiments) classification problems. Accuracy values have been recorded in the range of 95.5–99.8% for the two-class classification problem and 92–97.3% for the three-class classification problem. The LDA feature reduction outperformed PCA by an average of 4.34% and 3.52% in accuracy and F1 Score measures, respectively. The overall approach outperformed the existing related works in literature by far of 23% and 34% for accuracy and F1 Score, respectively. The experimental studies and the obtained results show the efficiency of the proposed solution, which employs optimized features that rely on integrating a feature reduction module, together with a well representative feature set based on negation handled triple combination of N-gram features and positive/negative N-grams counts features. The overall results demonstrate great improvement with 24% increase in accuracy, 93% savings in the feature space, and 97% decrease in the classification execution time.

A Hybrid Feature Selection Method RFSTL for Manufacturing Quality Prediction Based on a High Dimensional Imbalanced Dataset

Under Industry 4.0, manufacturing quality prediction has been gaining increased interest from researchers and manufacturers. From the analysis of previous studies on quality predictions using machine learning, it became clear that the high dimensionality and imbalance of data are major and common problems affecting the learning performance. This work uses a hybrid method to address this issue, applying a Synthetic Minority Oversampling Technique & TomekLinks balancing approach to create balanced data and using Random Forest as the feature selecting measurement to reduce the dimensionality of data. In addition, a Fine Gaussian Support Vector Machine (Fine Gaussian SVM) based on the representative set of features selected by the hybrid method utilized is employed in this work to predict product quality. The results of experimentation demonstrate that the hybrid method proposed in this work performs well for manufacturing quality prediction and offers a simple, quick and powerful way to address the problem of feature selection encountered by the imbalanced classification.

Investigating Deep Learning Based Breast Cancer Subtyping Using Pan-Cancer and Multi-Omic Data.

Breast Cancer comprises multiple subtypes implicated in prognosis. Existing stratification methods rely on the expression quantification of small gene sets. Next Generation Sequencing promises large amounts of omic data in the next years. In this scenario, we explore the potential of machine learning and, particularly, deep learning for breast cancer subtyping. Due to the paucity of publicly available data, we leverage on pan-cancer and non-cancer data to design semi-supervised settings. We make use of multi-omic data, including microRNA expressions and copy number alterations, and we provide an in-depth investigation of several supervised and semi-supervised architectures. Obtained accuracy results show simpler models to perform at least as well as the deep semi-supervised approaches on our task over gene expression data. When multi-omic data types are combined together, performance of deep models shows little (if any) improvement in accuracy, indicating the need for further analysis on larger datasets of multi-omic data as and when they become available. From a biological perspective, our linear model mostly confirms known gene-subtype annotations. Conversely, deep approaches model non-linear relationships, which is reflected in a more varied and still unexplored set of representative omic features that may prove useful for breast cancer subtyping.

Evaluating Solid-Electrolyte Interphases for Lithium and Lithium-free Anodes from Nanoindentation Features

Summary The morphology and mechanical property of SEIs for lithium anodes crucially affect the electrochemical cycling performance of Li-metal batteries such as Li-S and Li-O2 batteries. Herein, we establish the standards for rapidly assessing SEIs for Li and Li-free type of anodes by AFM nanoindentation features toward prediction of corresponding electrochemical performances. A series of single-layered and multi-layered SEIs with known composition and structures are purposely constructed. A set of unique nanoindentation features representative of mechanical properties of the basic SEIs are formed, serving as the standards for rapidly assessing the mechanical properties and electrochemical performances of unknown SEIs together with their Young’s modulus, smoothness, and thickness. To validate the applicability of the method, SEIs formed by electrochemical aging and formed on Cu substrates are further evaluated. Our work not only contributes to understanding the influence of SEIs on cycling performances but also provides a rapid way for screening SEIs.

A multi-objective artificial butterfly optimization approach for feature selection

Feature selection plays an essential role in machine learning since high dimensional real-world datasets are becoming more popular nowadays. The very basic idea consists in selecting a compact but representative set of features that reduce the computational cost and minimize the classification error. In this paper, the authors propose single, multi- and many-objective binary versions of the Artificial Butterfly Optimization (ABO) in the context of feature selection. The authors also propose two different approaches: (i) the first one (MO-I) aims at optimizing the classification accuracy of each class individually, while (ii) the second one (MO-II) considers the feature set minimization in the process either. The experiments were conducted over eight public datasets, and the proposed approaches are compared against the well-known Particle Swarm Optimization, Firefly Algorithm, Flower Pollination Algorithm, Brainstorm Optimization, and the Black Hole Algorithm. The results showed that the binary single-objective ABO performed better than the other meta-heuristic techniques, selecting fewer features and also figuring a lower computational burden. Concerning multi- and many-objective feature selection, both MO-I and MO-II approaches performed better than their single-objective meta-heuristic counterparts.

Anomaly Detection of Manufacturing Process for Multi-Variety and Small Batch Production

A series of quality control problems arise in multi-variety and small batch production due to the lack of data and other factors. In this paper, an integrated model in which control chart pattern (CCP) recognition is applied for anomaly detection is proposed to solve the problems. The integrated model is made up of four parts: feature extraction module, feature selection module, classifier module and anomaly diagnosis module. In the first module, thirteen shape features and eight statistical features of control charts are extracted. In the second module, the most representative feature set is selected by the sequential floating forward selection (SFFS) method. In the third module, a multiclass support vector machine (MSVM) which is optimized by beetle antennae search (BAS) algorithm is used to identify abnormal CCPs. In the last module, the results of pattern recognition are utilized to analyze the possible causes. The simulation results show that the CCP recognition method proposed in this paper has higher classification accuracy than other competing methods in the case of small sample with small amount of data. Finally, an example verifies that the proposed anomaly detection method is effective in multi-variety and small batch manufacturing environment.