Search

The further scientific applications of DMSP-OLS night-time light (NTL) imagery have been being limited by the accuracy, automation, and speed of radiometric intercalibration. In order to solve the aforementioned problems, this article is the first to propose a new least-median-of-squares (LMedS)-based power regression (LBPR) for automatically radiometric intercalibration and investigate the reasons for the optimal model of radiometric intercalibration, especially those based on the Taylor expansion and probability principle. NTL data in six regions all over the world, from 1994 and 1997 to 2007, were used as the test datasets. When the five kinds of LMedS-based radiometric intercalibration models (i.e., linear, quadratic, power, exponential, and logarithmic regression) are synthetically compared in absolute accuracy (adjusted RMSE) and running speed, it is concluded that the LBPR, which has the highest accuracy and preferable running speed, is recommended as the optimal method, which can also be used as a reference for other types of imagery preprocessing.

Air-writing recognition has received wide attention due to its potential application in intelligent systems. To date, some of the fundamental problems in isolated writing have not been addressed effectively. This paper presents a simple yet effective air-writing recognition approach based on deep convolutional neural networks (CNNs). A robust and efficient hand tracking algorithm is proposed to extract air-writing trajectories collected by a single web camera. The algorithm addresses the push-to-write problem and avoids restrictions on the users’ writing without using a delimiter and an imaginary box. A novel preprocessing scheme is also presented to convert the writing trajectory into appropriate forms of data, making the CNNs trained with these forms of data simpler and more effective. Experimental results indicate that the proposed approach not only obtains much higher recognition accuracy but also reduces the network complexity significantly compared to the popular image-based methods.

In recent times, people have become increasingly health-conscious. To obtain timely and accurate information on the status of the heart, one of the most important organs of the human body, there is a growing demand among individuals and doctors for accurate and real-time automatic classification of arrhythmias. Consequently, this paper proposes a fast and accurate classification method for arrhythmias. In the proposed method, we build an incremental broad learning (IBL) classification model based on the biased dropout technique for arrhythmia-type recognition. In particular, we extract the morphological-rhythm features of the denoised signal as the input data of the IBL model in the electrocardiogram signal preprocessing. The IBL model enhances the classification effect of the node optimization model by using improved features. To the best of our knowledge, this study is the first application of the IBL model to the study of arrhythmia classification. The results of experiments conducted on the MIT-BIH database indicate that the proposed method is effective and achieves superior classification results. The average classification accuracy for six types of arrhythmias was 99%, and the training time required was only 2.7 s. In addition, based on the evaluation index recommended by the ANSI/AAMI EC57:2012 standard, our method is superior to existing methods on all indexes, except for the positive predictive rate of ventricular ectopic beats. Therefore, the proposed classification method outperforms state-of-the-art methods in terms of real-time performance and accuracy and provides a new approach for further improvements in arrhythmia classification.

In recent years, with the rise of environmental awareness worldwide, the number of solar power plants has significantly increased. However, the maintenance of solar power plants is not an easy job, especially the detection of malfunctioning photovoltaic (PV) cells in large-scale or remote power plants. Therefore, finding these cells and replacing them in time before severe events occur is increasingly important. In this paper, we propose a hybrid scheme with three embedded learning methods to enhance the detection of malfunctioning PV modules with validated efficiencies. For the first method, we combine the improved gamma correction function (preprocess) with a convolutional neural network (CNN). Infrared (IR) thermographic images of solar modules are used to train the abovementioned improved algorithm. For the second method, we train a CNN model using the IR temperatures of PV modules with the preprocessing of a threshold function. A compression procedure is then designed to cut the time-consuming preprocesses. The third method is to replace the CNN with the eXtreme Gradient Boosting (XGBoost) algorithm and the selected temperature statistics. The experimental results show that all three methods can be implemented with high detection accuracy and low time consumption, and furthermore, the hybrid scheme provides an even better accuracy.

Polarimetric synthetic aperture radar (PolSAR) imagery can provide more intuitive and detailed SAR polarization information, and it is widely used in the classification and semantic segmentation of remote sensing. To bridge the PolSAR data and application, the 2020 Gaofen Challenge on Automated High-Resolution Earth Observation Image Interpretation provides a set of high-quality PolSAR semantic segmentation dataset. A series of preprocessing methods is first used to analyze the PolSAR images to improve the semantic segmentation performance of the PolSAR imagery. A special polarimetric decomposition method is used to extract the features, and the filter and the data truncation are implemented to enhance local and global information of images. And the random region matting method is proposed to expand the training samples. Finally, the DeepLabV3+ method with the ResNet101-V2 is employed to achieve the semantic segmentation. A variety of comparison experiments verifies the effectiveness of our methods. Simultaneously, compared with the classification methods of other groups in the competition, our methods have obvious advantages in the inference time and semantic segmentation accuracy. The proposed method achieved a frequency weighted intersection over union of 75.29% in the contest.

Recently, development in intelligent transportation systems (ITS) requires the input of various kinds of data in real-time and from multiple sources, which imposes additional research and application challenges. Ongoing studies on Data Fusion (DF) have produced significant improvement in ITS and manifested an enormous impact on its growth. This paper reviews the implementation of DF methods in ITS to facilitate traffic flow analysis (TFA) and solutions that entail the prediction of various traffic variables such as driving behavior, travel time, speed, density, incident, and traffic flow. It attempts to identify and discuss real-time and multi-sensor data sources that are used for various traffic domains, including road/highway management, traffic states estimation, and traffic controller optimization. Moreover, it attempts to associate abstractions of data level fusion, feature level fusion, and decision level fusion on DF methods to better understand the role of DF in TFA and ITS. Consequently, the main objective of this paper is to review DF methods used for real-time and multi-sensor (heterogeneous) TFA studies. The review outcomes are (i) a guideline of constructing DF methods which involve preprocessing, filtering, decision, and evaluation as core steps, (ii) a description of the recent DF algorithms or methods that adopt real-time and multi-sensor sources data and the impact of these data sources on the improvement of TFA, (iii) an examination of the testing and evaluation methodologies and the popular datasets and (iv) an identification of several research gaps, some current challenges, and new research trends.

Aircraft recognition in high-resolution remote sensing images has rapidly progressed with the advance of convolutional neural networks (CNNs). However, the previous CNN-based methods may not work well for recognizing aircraft in low-resolution remote sensing images because the blurred aircraft in these images offer insufficient details to distinguish them from similar types of targets. An intuitive solution is to introduce superresolution preprocessing. However, conventional superresolution methods mainly focus on reconstructing natural images with detailed texture rather than constructing a high-resolution object with strong discriminative information for the recognition task. To address these problems, we propose a unified framework for joint superresolution and aircraft recognition (Joint-SRARNet) that tries to improve the recognition performance by generating discriminative, high-resolution aircraft from low-resolution remote sensing images. Technically, this network integrates superresolution and recognition tasks into the generative adversarial network (GAN) framework through a joint loss function. The generator is constructed as a joint superresolution and refining subnetwork that can upsample small blurred images into high-resolution ones and restore high-frequency information. In the discriminator, we introduce a new classification loss function that forces the discriminator to distinguish between real and fake images while recognizing the type of aircraft. In addition, the classification loss function is back-propagated to the generator to obtain high-resolution images with discriminative information for easier recognition. Extensive experiments on the challenging multitype aircraft of remote sensing images (MTARSI) dataset demonstrate the effectiveness of the proposed method in restoring a clear super-resolved image from a small blurred image and significant improvement in the recognition performance. To our knowledge, this is the first work on joint superresolution and aircraft recognition tasks.

Millimeter-wave is widely used for consumer radar applications like driver assistance systems in automated vehicles and gesture recognition in touch-free interfaces. To cope with the increased hardware complexity, higher costs and power consumption of wideband systems at millimeter-wave frequencies, we propose a fully digital architecture with low-resolution analog-to-digital converters (ADCs) on each radio-frequency chain. The effect of the low-resolution ADCs on radar parameter estimation is characterized by the Cramér-Rao bound (CRB) under the proposed hardware constraints. Prior work has shown that at low signal-to-noise ratio, a radar system with 1-bit ADCs suffers a performance loss of 2 dB in parameter estimation compared to a system with ideal infinite resolution ADCs. In this paper, we design an analog preprocessing unit that beamforms in a particular direction and improves the system performance in terms of the achievable CRB. We optimize the proposed preprocessing architecture and show that the optimized network is realizable through low-cost low-resolution phase-shifters. With the optimized preprocessor network in the system, we reduce the gap to 1.16 dB compared to a system with ideal ADCs. We demonstrate the potential of the proposed architecture to meet the requirements of high-resolution sensing through analytical derivation and numerical computation of an improved CRB and show its achievability through a correlation-based estimator.

Imbalanced learning problems are a challenge faced by classifiers when data samples have an unbalanced distribution in each class. Furthermore, the synthetic oversampling method (SMOTE) is a preprocessing technique widely used to synthesize new data and balance the different numbers of samples in each class. One of the SMOTE method’s expansions is based on the initial selection approach, which determines the best candidates to be oversampled in the data before the process of synthetic example generation starts. However, SMOTE and most of the existing oversampling methods based on initial selection still found overlapping data on the final result. This issue makes it difficult for any classifiers to determine the decision boundary of each class. Therefore, this research proposes a new oversampling technique called Radius-SMOTE, which emphasizes the initial selection approach by creating synthetic data based on a safe radius distance. Furthermore, new synthetic data are prevented from overlapping in the opposite class with the safe radius distance. The Radius-SMOTE was evaluated extensively with thirteen artificial imbalanced datasets from the KEEL repository. The experimental results show that the proposed method is able to achieve the best results on 5 datasets, namely yeast-1-4-5-8_vs_7, ecoli-0-1-3-7_vs_2-6, Umbilical cord, Pima, and Haberman dataset in term of various assessment metrics. Besides that, the computational cost for our proposed method is also relatively low, with an average time of 0.5 to 1 second on the 13 tested datasets.

Due to the rapid developments in technology and the sudden expansion of social media use, Dialect Arabic has become an important source of data that needs to be addressed when building Arabic corpora. In this paper, thirty-three Arabic corpora are surveyed to show that despite all of the developments in the literature, Saudi dialect (SD) corpora still need further expansion. This paper contributes to the literature on SD corpora by creating the largest Saudi corpus – the King Saud University Saudi Corpus (KSUSC) – with +1B total words, including +119M SD words. The KSUSC not only is the newest and largest SD corpus but is also diverse, covering 26 domains in text collected from five different sources. This paper also contributes to the literature by developing a new incremental preprocessing system that is used to create relevant lexicons that are then used to clean and normalize the collected data. This incremental system is scalable and can be adapted for different resources and dialects. Moreover, the collection process for building the KSUSC is discussed in detail, and the challenges in collecting SD text with respect to each platform are highlighted. By the end of this paper, different design criteria are proposed and used with the KSUSC to conclude that the resulting corpus can be of great benefit to researchers who are interested in integrating the corpus with their own work or using its resulting lexicons with Saudi-based NLP tasks.

Pansharpening aims at fusing a low-resolution multispectral (MS) image and a high-resolution (HR) panchromatic (PAN) image acquired by a satellite to generate an HR MS image. Many deep learning based methods have been developed in the past few years. However, since there are no intended HR MS images as references for learning, almost all of the existing methods downsample the MS and PAN images and regard the original MS images as targets to form a supervised setting for training. These methods may perform well on the down-scaled images; however, they generalize poorly to the full-resolution images. To conquer this problem, we design an unsupervised framework that is able to learn directly from the full-resolution images without any preprocessing. The model is built based on a novel generative multiadversarial network. We use a two-stream generator to extract the modality-specific features from the PAN and MS images, respectively, and develop a dual discriminator to preserve the spectral and spatial information of the inputs when performing fusion. Furthermore, a novel loss function is introduced to facilitate training under the unsupervised setting. Experiments and comparisons with other state-of-the-art methods on GaoFen-2, QuickBird, and WorldView-3 images demonstrate that the proposed method can obtain much better fusion results on the full-resolution images. Code is available. [Online]. Available: https://github.com/zhysora/PGMAN.

Human action recognition techniques have gained significant attention among next-generation technologies due to their specific features and high capability to inspect video sequences to understand human actions. As a result, many fields have benefited from human action recognition techniques. Deep learning techniques played a primary role in many approaches to human action recognition. The new era of learning is spreading by transfer learning. Accordingly, this study’s main objective is to propose a framework with three main phases for human action recognition. The phases are pre-training, preprocessing, and recognition. This framework presents a set of novel techniques that are three-fold as follows, (i) in the pre-training phase, a standard convolutional neural network is trained on a generic dataset to adjust weights; (ii) to perform the recognition process, this pre-trained model is then applied to the target dataset; and (iii) the recognition phase exploits convolutional neural network and long short-term memory to apply five different architectures. Three architectures are stand-alone and single-stream, while the other two are combinations between the first three in two-stream style. Experimental results show that the first three architectures recorded accuracies of 83.24%, 90.72%, and 90.85%, respectively. The last two architectures achieved accuracies of 93.48% and 94.87%, respectively. Moreover, The recorded results outperform other state-of-the-art models in the same field.

Magnetic resonance imaging (MRI) is a medical technology that uses powerful magnets, radio waves, and a computer to produce images of the body’s internal organs. The patient should be quiet and motionless during the scanning period, as unavoidable movements, such as breathing and heart rate, cause motion artifacts in the image, which cause contrast instability and low-resolution MRI images. Imaging in the clinical setting is performed at low resolution because the scanning time for high-resolution MR imaging would be very long and cumbersome, and imaging is also very expensive. Learning-based image superresolution methods can reconstruct MR images with optimal quality and resolution. However, these methods have problems such as the inability to find the intrinsic relationship between low- and high-resolution image patches in the training dictionary, specification of the proper amount of error in the training and testing stage due to variability in MR image contrast, inability to reconstruct objects by smoothed edges, and use of the backpropagation method in updating their weight vectors. In this paper, we propose a new superresolution method according to competitive learning-based approaches to overcome the problems of previous superresolution methods, which do not have the problems and complexities of those methods. The proposed method includes self-organizing maps as a preprocessor, the nearest neighbor algorithm as a classifier, and a high-frequency filter as a high-frequency image detail extractor. We constructed a single external dictionary from a combination of low-resolution and high-resolution feature patches and trained our SOM network. Next, we reconstruct the high-resolution image by converting the low-resolution input image into feature patch vectors, and for each vector, we find all corresponding neurons in the network and retrieve all their training feature vectors. Next, we train the nearest neighbor algorithm with the recovered vectors plus the input vector and find the best similarity vector to the input vector. After finding all the best similarity vectors to the input vectors, we reconstruct our high superresolution image. The proposed image superresolution method in practical experiments was trained, tested, and evaluated by the Div2k dataset and compared with other traditional and state-of-the-art image enhancement methods on various datasets, such as SET5, SET14, BSDS100, and URBAN100, and presented better results with higher accuracy and quality than traditional and state-of-the-art methods, both visually compared to each other by human and computational benchmarks, such as the peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM), to compare image superresolution algorithms. This method is best for reconstructing high-resolution magnetic resonance images that require high-frequency details and sharp edges with a smooth slope of the imaging objects in their structures. The execution time of the proposed algorithm is slower than that of the other algorithms, so we use GPU hardware and parallel programming techniques to increase the algorithm speed.

Recently, deep learning models have emerged as promising methods for the diagnosis of different diseases. Cardiac disease is among the leading life-threatening diseases on a global scale. The aim of this paper is to propose an optimized Convolutional Neural Network (CNN) model for the classification of electrocardiogram (ECG) heartbeat data. The proposed ECG classification approach is designed with an optimal CNN configuration to classify cardiac arrhythmias quickly and effectively. Finding an optimal configuration for the CNN hyperparameters is time-consuming and needs extensive experimentation. To overcome this challenge, we present an optimization step for the proposed CNN model using a customized genetic algorithm. It provides an automatic suggestion for the best hyperparameter settings of the proposed CNN. The challenge in utilizing the genetic algorithm is that its operators need to be customized to handle our problem domain. Our approach accepts raw ECG signals without any preprocessing steps, which has benefit in saving the computation time. Our approach also provides a resampling step to ensure generalization, to better handle imbalanced ECG classes. Experiments show promising results of our proposed approach against other approaches whose CNN hyperparameters setting depended on numerous trials, requiring extensive ECG feature extraction steps, and do not consider imbalanced classes. The performance of our proposed approach is better than other existing methods both in terms of higher classification accuracy (98.45%), and lower computational complexity.

This article proposes boosting the multiplication performance for convolutional neural network (CNN) inference using a precision prediction preprocessor which controls various precision approximate multipliers. Previously, utilizing approximate multipliers for CNN inference was proposed to enhance the power, speed, and area at a cost of a tolerable drop in the accuracy. Low precision approximate multipliers can achieve massive performance gains; however, utilizing them is not feasible due to the large accuracy loss they cause. To maximize the multiplication performance gains while minimizing the accuracy loss, this article proposes using a tiny two-class precision controller to utilize low and high precision approximate multipliers hybridly. The performance benefits for the proposed concept are presented for multi-core multi-precision architectures and single-core reconfigurable architectures. Additionally, a design for a merged reconfigurable approximate multiplier with two precisions is proposed for utilization in single-core architectures. For performance comparison, several segments-based approximate multipliers with different precisions were synthesized using CMOS 15nm technology. For accuracy evaluation, the concept was simulated on VGG19, Xception, and DenseNet201 using the ImageNetV2 dataset. This article will demonstrate that the proposed concept can achieve significant performance gains with a minimal accuracy loss when compared to designs that utilize exact multipliers or single-precision approximate multipliers.

Epilepsy is one of the world’s most common neurological diseases. Reliable early prediction and warning of seizures can provide timely treatment for patients with epilepsy, and improve their quality of life. Compared with most hand-designed prediction methods, an automatic prediction model that can process the original electroencephalogram (EEG) signals directly and take into account the leads optimization problem is needed. In this paper, we proposed an end-to-end automatic seizure prediction model based on the Batch Normalization Long Short Term Memory networks (BNLSTM) and Channel and Spatial attention (CASA). Firstly, raw EEG signals without any preprocessing are used as the input to the system, which can reduce the computation amount. Secondly, BNLSTM and CASA retained the time and spatial information of the raw EEG data respectively. Channel attention (CA) achieved the automatic optimization of EEG full-lead data and improved the prediction accuracy. Spatial attention (SA) achieved the adaptive learning of feature parameters. Finally, a fully connected layer is applied to predict the seizures. The performance of the seizure prediction model we proposed is evaluated on the data of 14 patients with Area Under the Curve (AUC) of 0.986, accuracy (Acc) of 0.956, specificity (Spe) of 0.968, and sensitivity (Sen) of 0.942. In addition, the proposed method provided an accurate prediction for all 50 seizures of the other 5 patients in the generalization dataset. Experimental results show that the proposed model has a certain generalization performance, which can provide a reliable basis for early warning of epileptic seizures.

Gastric cancer is the third-most-common cause of cancer-related deaths in the world. Fortunately, it can be detected using endoscopy equipment. Computer-aided diagnosis (CADx) systems can help clinicians identify cancer from gastric diseases more accurately. In this paper, we present a CADx system that distinguishes and classifies gastric cancer from pre-cancerous conditions, such as gastric polyps, gastric ulcers, gastritis, and bleeding. The system uses a deep-learning model, Xception, which involves depth-wise separable convolutions, to classify cancer and non-cancers. The proposed method consists of two steps: Google’s AutoAugment for augmentation and the simple linear iterative clustering (SLIC) superpixel and fast and robust fuzzy C-means (FRFCM) algorithm for image segmentation during preprocessing. These approaches produce a feasible method of distinguishing and classifying cancers from other gastric diseases. Based on biopsy-supported ground truth, the performance metrics of the area under the receiver operating characteristic curve (i.e. Az) are measured on the test sets. Based on the classification results, the Az of the proposed classification model is 0.96, which is 0.06 up from 0.90 which is the Az of the original data. Our methods are fully automated without the manual specification of region-of-interests for the test and with a random selection of images for model training. This methodology may play a crucial role in selecting effective treatment options without the need for a surgical biopsy.

This paper is based on an improved three-dimensional U-net convolutional neural network deep learning algorithm for heart coronary artery segmentation for disease risk prediction, and it is practical with multiple data sets under two backgrounds without centerline and with the centerline. By using a new local feature to extract the ventricular information, and using the deep belief network to extract the features to regress the contour coordinates of the biventricular. Combining features and deep belief networks and training regression networks can not only extract high-level information but also accurately divide the left and right ventricles at a small computational cost. The performance of segmentation based on the dice coefficient compared between the two datasets. The results show that the model training effect of the centerline preprocessing is superior to the original data. The experimental results show that the best effect reaches the dice coefficient of 0.8291. In the experiment, it found that simple data expansion may be detrimental to the test data. From the training curve, it is believed that with the improvement of the quality of training data, the performance of coronary artery segmentation can be further improved, and it is of great significance to provide doctors and patients with more accurate and efficient opinions and suggestions in clinical practice to improve the quality of diagnosis and treatment. The purpose of assisting experts in real-time diagnosis and analysis achieved.

Limited-angle computed tomography (CT) has arisen in some medical and industrial applications. It is also a challenging problem since some scan views are missing and the directly reconstructed images often suffer from severe distortions. For such kind of problems, we analyze the features of limited-angle CT images and propose a multi-scale dilated convolution neural network (MSD-CNN) to correct the artifacts and to restore the image. In this network, the dilated convolution layer and multi-scale pooling layer are combined to form a group and exited in the whole encoder-decoder process. Since the dilated convolutions support an exponential expansion of the receptive field without losing resolution and coverage, the obtained artifact features possess the multi-scale characteristic. Furthermore, to improve the effectiveness and accuracy of the training step, we employ a preprocessing method, which extracts image patches. Numerical experiments verify the out-performance of the proposed method compared with some conventional methods, such as Unet based deep learning,TV- and L0 -based optimization methods.

Traveling salesman problem (TSP) has been widely applied to various fields of production and life. In order to solve the TSP, this work optimizes the architecture of the particle swarm optimization (PSO) by introducing a preprocessing multi-subdomain grouping approach to divide the entire search space into several sub-regions, based on the geographical center of planned cities. To solve subdomain problems, the global search ability of PSO is further improved using genetic mutation and particle competition mechanism to maintain the population diversity during late iterations. An effective simplified 4-opt is finally employed to eliminate path-crossing after connecting all subdomain paths in a time saving way. Comparison with three of our previous algorithms and seven algorithms from other reports reveals that the proposed algorithm shows good performance in both the computing efficiency and route quality. Particularly, it is suitable for small (or medium)-sized TSP problems with relatively percentage error values below 4.8%.

Most hyperspectral anomaly detection algorithms are based on various hypothetical models justified by different methods. The closer to the real-world scene distribution a hypothetical model is, the better detection performance usually results, albeit at the expense of increased model complexity. There is also a challenge in most of the detection methods that anomalous components cannot be completely separated from the background due to the difference between hypothetical models and the real-world scene. To address this problem, a novel strategy based on graph theory and the multiple support vector machines technique is proposed in this paper. Firstly, a graphical connected point-based estimation is utilized as a preprocessing procedure to separate potential anomalies from robust backgrounds of the image. Second, without building a hypothetical model, the hyperspectral image is classified into several categories by a clustering method according to the different characteristics of materials. Consequently, for each category, the support vector machine is used to classify anomalous components and background components belonging to specific categories, and the result map of each classification is obtained. Finally, all maps corresponding to various categories are fused by an effective fusion strategy, and the detection result is generated by our method. In the subsequent experiments, we use several state-of-art anomaly detection algorithms in a comparison with our method using both simulated and real-world HSI datasets. The experimental results demonstrate that the proposed method outperforms others.

Compared to conventional photography which allows capturing spatial intensity only, light field imaging can capture angular and spatial information by collecting information from all directions. This massive information could be used in many applications such as depth estimation and post-capture refocusing. In addition, the emergence of consumer light field cameras has resulted in the widespread use of light field imaging. However, its limited resolution constitutes a major drawback to the use of enormous capabilities provided. In this article, we tried to alleviate the effect of this drawback by using a machine learning algorithm. Our proposed network is built upon existing reconstruction techniques that divide the process into disparity estimation and final image reconstruction. This is achieved by using two consecutive neural networks while the whole network is trained simultaneously. We propose to use a predefined convolutional network at the first stage to decrease preprocessing time, and in addition, we use dual disparity vectors to alleviate the interpolation error whereas warping input images. Our system was trained to reconstruct multi-angular resolution images fast and accurately using real light field images for training. Experimental results demonstrate that the proposed system can reconstruct high-quality images faster than the state-of-the-art techniques.

Duplicate bug report detection is a process of finding a duplicate bug report in the bug tracking system. This process is essential to avoid unnecessary work and rediscovery. In typical bug tracking systems, more than thousands of duplicate bug reports are reported every day. In turn, human cost, effort and time are increased. This makes it an important problem in the software management process. The solution is to automate the duplicate bug report detection system for reducing the manual effort, thus the productivity of triager’s and developer’s is increased. It also speeds up the process of software management as a result software maintenance cost is also reduced. However, existing systems are not quite accurate yet, in spite of these systems used various machine learning approaches. In this work, an automatic bug report detection and classification model is proposed using deep learning technique. The proposed system has three modules i.e. Preprocessing, Deep Learning Model and Duplicate Bug report Detection and Classification. Further, the proposed model used Convolutional Neural Network based deep learning model to extract relevant feature. These relevant features are used to determine the similar features of bug reports. Hence, the bug reports similarity is computers through these similar features. The performance of the proposed system is evaluated on six publicly available datasets using six performance metrics. It is noticed that the proposed system outperforms the existing systems by achieving an accuracy rate in the range of 85% to 99 % and recall@k rate in between 79%–94%. Moreover, the effectiveness of the proposed system is also measured on the cross training datasets of same and different domain. The proposed system achieves a good high accuracy rate for same domain data sets and low accuracy rate for different domain datasets.

Satellite images are primary data in weather prediction modeling. Deep learning-based approach, a viable candidate for automatic image processing, requires large sets of annotated data with diverse characteristics for training purposes. Accuracy of weather prediction improves with data having a relatively dense temporal resolution. We have employed interpolation and data augmentation techniques for enhancement of the temporal resolution and diversifications of characters in a given dataset. Algorithm requires classical approaches during preprocessing steps. Three optical flow methods using 14 different constraint optimization techniques and five error estimates are tested here. The artificially enriched data (optimal combination from the previous exercise) are used as a training set for a convolutional neural network to classify images in terms of storm or nonstorm. Several cyclone data (eight cyclone datasets of a different class) were used for training. A deep learning model is trained and tested with artificially densified and classified storm data for cyclone classification and locating the cyclone vortex giving minimum 90% and 84% accuracy, respectively. In the final step, we show that the linear regression method can be used for predicting the path.

Automatic detection of the pectoral muscle in mammograms is widely used in computer-aided diagnostic (CAD) systems for breast cancer. The pectoral muscle region has some prominent features such as the upper corner position, high density, and triangular shape. But, these features may be distorted due to the masses, artifacts, skin folds, and overlapping tissues, and other reasons. Despite recent developments in CAD technology, accurate detection of distorted pectoral muscle images remains a challenging task. In this study, we proposed an automatic method that uses a divided topographic representation to detect distorted pectoral muscle boundaries. After the preprocessing stage, firstly an isocontour map is generated and then divided into horizontal blocks. The contours of the pectoral muscle boundary in the blocks often reveal specific patterns in terms of location, geometric and topological features. We developed a new segmentation algorithm, rule-based contour detection (RBCD), to detect these specific patterned isocontours. The method applied to two datasets consisting of 84 and 201 mammogram images from MIAS and Inbreast databases respectively. Besides, some distorted pectoral muscle samples selected from these datasets were used to further analyze the performance of the proposed method. The mean False-Positive and the mean False-Negative rates of the proposed method for MIAS and Inbreast datasets were 0.92%, 1.26%, and 2.34%, 1.15%, respectively. The quantitative and qualitative results for the distorted pectoral muscle samples show that the proposed method outperformed the compared methods.

Airborne laser scanning (ALS) data is one of the most commonly used data for terrain products generation. Filtering ground points is a prerequisite step for ALS data processing. Traditional filtering methods mainly use handcrafted features or predefined classification rules with preprocessing/post-processing operations to filter ground points iteratively, which is empirical and cumbersome. Deep learning provides a new approach to solve classification and segmentation problems because of its ability to self-learn features, which has been favored in many fields, particularly remote sensing. In this article, we proposed a point-based fully convolutional neural network (PFCN) which directly consumed points with only geometric information and extracted both point-wise and tile-wise features to classify each point. The network was trained with 37449157 points from 14 sites and evaluated on 6 sites in various forested environments. Additionally, the method was compared with five widely used filtering methods and one of the best point-based deep learning methods (PointNet++). Results showed that the PFCN achieved the best results in terms of mean omission error (T1 = 1.10%), total error (Te = 1.73%), and Kappa coefficient (93.88%), but ranked second for the root mean square error of the digital Terrain model caused by the worst commission error. Additionally, our method was on par with or even better than PointNet++ in accuracy. Moreover, the method consumes one-third of the computational resource and one-seventh of the training time. We believe that PFCN is a simple and flexible method that can be widely applied for ground point filtering.

Recently, agricultural remote sensing community has endeavored to utilize the power of artificial intelligence (AI). One important topic is using AI to make the mapping of crops more accurate, automatic, and rapid. This article proposed a classification workflow using deep neural network (DNN) to produce high-quality in-season crop maps from Landsat imageries for North Dakota. We use historical crop maps from the agricultural department and North Dakota ground measurements as training datasets. Processing workflows are created to automate the tedious preprocessing, training, testing, and postprocessing workflows. We tested this hybrid solution on new images and received accurate results on major crops such as corn, soybean, barley, spring wheat, dry beans, sugar beets, and alfalfa. The pixelwise overall accuracy in all three test regions is over 82% for all land types (including noncrop land), which is the same level of accuracy as the U.S. Department of Agriculture Cropland Data Layer. The texture of DNN maps is more consistent with fewer noises, which is more comfortable to read. We find DNN is better on recognizing big farmlands than recognizing the scattered wetlands and suburban regions in North Dakota. The model trained on multiple scenes of multiple years and months yields higher accuracy than any of the models trained only on a single scene, a single month, or a single year. These results reflect that DNN can produce reliable in-season maps for major crops in North Dakota big farms and could provide a relatively accurate reference for the minor crops in scattered wetland fields.

The existing approaches usually treat the recognition of pornographic images as a binary or multi-class classification task, which is realized by extracting a variety of features. However, these approaches ignore the problem of infinite kinds in the negative samples and lose sight of the impact of uncertainty in classification tasks, resulting in inadequate negative data sets and incorrect recognition of samples that are not in the training set. In order to address this challenge, this paper proposes a method named Deep One-Class with Attention for Pornography (DOCAPorn) that recognizes the pornographic images through the one-class classification model based on neural networks and introduces the visual attention mechanism to enhance the performance of recognition. Moreover, since the existing approaches based on deep convolutional neural networks (CNNs) require a fixed-size (e.g., $224\times224$ ) input image, the geometric distortion caused by image scaling is not considerd in the existing approaches, which reduces the pornographic image recognizition accuracy. In order to solve this issue, this paper proposes the Scale Constraint Pooling (SCP) that converts the inputs of different dimensions into outputs of the same dimension. In addition, all the existing approaches ignore the adversarial attacks in the field of pornographic image recognition. In order to deal with this problem, the paper proposes the Preprocessing for Compressing and Reconstructing (PreCR), a pre-processing approach that reduces the subtle perturbation through compressing the images and then reconstructs the purified image for recognition. The proposed approach is verified by conducting comparative experiments using custom datasets. The experimental results showed that we achieved an accuracy of 98.419% on our dataset. In addition, the proposed approach yielded a recognition accuracy of 95.632% on the NPDI dataset. Furthermore, the obtained results demonstrate that the proposed approach not only effectively recognizes pornographic images but also effectively defends the adversarial attacks.

Image inpainting is one of the most important problems in the field of image algorithm, and it is also an effective preprocessing method for many other image applications. In this paper, we suggest an image decontamination method which is mainly used to remove mesh stains and also provide a data set for this task. To our knowledge, this work is the first attempt to solve this kind of problem. Specifically, the proposed method is composed of two phases: we first remove the mesh stains with an Improved Conditional Generative Adversarial Network, and then utilize a Semantic Network to fine tune the details. Experiments demonstrated that this two-stage model can remove the mesh stains. Results show that our method significantly out-performs existing methods and achieves superior inpainting results on challenging cases.

Compared with the state-of-the-art architectures, using the 3D point cloud as the input of the 2D convolutional neural network without preprocessing will restrict the feature expression of the network. To address this issue, we propose a high-precision classification network using bearing angle (BA) images, depth images, and RGB images. Due to the development of unmanned vehicles, determining how to recognize objects from the information collected by sensors is important. Our approach takes data from LiDAR and a camera and projects a 3D point cloud into 2D BA images and depth images. The RGB image captured by the camera is used to select the region of interest (ROI) corresponding to the point cloud. However, only adding input information is not enough to improve the classification ability of general convolutional neural networks. In our approach, we use a Dense-Residual Fusion Network (DRF-Net), which consists of Dense-Residual Blocks (DRBs). The Dense-Residual Fusion Network can achieve 97.92% accuracy with three input formats on a KITTI raw dataset.

Liver and liver tumor segmentation provides vital biomarkers for surgical planning and hepatic diagnosis. In this paper, we propose and validate a novel level-set method integrating an enhanced edge indicator and an automatically derived initial curve for CT based liver tumor segmentation. At the preprocessing step, the CT image intensity values were truncated to lie in a fixed range to enhance the image contrast surrounding liver and liver tumor. To remove non-liver tissues for subsequent tumor segmentation, liver was firstly segmented using two convolutional neural networks in a coarse-to-fine manner. A 2D slice-based U-net was used to roughly localize the liver and a 3D patch-based fully convolutional network was used to refine the liver segmentation as well as to roughly localize the liver tumor. A novel level-set method was then presented to further refine the tumor segmentation. Specifically, the probabilistic distribution of the liver tumor was estimated using unsupervised fuzzy c-means clustering, which was then utilized to enhance the edge-detector used in level-set. Effectiveness of the proposed pipeline was validated on two publicly-available datasets. Experimental results identified the superior segmentation performance of the proposed pipeline over state-of-the-art methods.

Understanding and exploiting topographical data via standard machine learning techniques is challenging, mainly due to the large dynamic range of values present in elevation data and the lack of direct relationships between anthropogenic phenomena and topography, when considering topographic–geology couplings, for instance. Here we consider the first hurdle, dynamic range, in an effort to apply Convolutional Neural Network (CNN) approaches for prediction of human activity. CNN for learning 3-D elevation data relies on data normalization approaches, which only consider locally available points, thereby discarding contextual information and eliminating global contrast cues. We present a fully invertible and data-driven global partitioning elevation normalization (GPEN) preprocessing technique, which is intended to ameliorate the impact of limited data dynamic range. Global elevation populations are derived and used to formulate a distribution, which is used to adopt a partitioning scheme to remap all values according to global occurrence frequency, while preserving partition contrast. Using USGS 3-D Elevation Project and Microsoft building footprint data, we conduct a binary classification experiment predicting building footprint presence from elevation data, with and without a global remapping using the SegNet convolutional encoder-decoder model. The results of the experiment show more rapid model convergence, reduced regionalization errors, and enhanced classification metrics when compared to standard normalization preprocessing techniques. GPEN demonstrates performance over 10% higher than the next best conventional preprocessing method, with a mean overall accuracy of 94.76%. GPEN may show promise as an alternative normalization for deep learning with topological data.

Conventionally, as a preprocessing step, functional MRI (fMRI) data are spatially smoothed before further analysis, be it for activation mapping on task-based fMRI or functional connectivity analysis on resting-state fMRI data. When images are smoothed volumetrically, however, isotropic Gaussian kernels are generally used, which do not adapt to the underlying brain structure. Alternatively, cortical surface smoothing procedures provide the benefit of adapting the smoothing process to the underlying morphology, but require projecting volumetric data on to the surface. In this paper, leveraging principles from graph signal processing, we propose a volumetric spatial smoothing method that takes advantage of the gray-white and pial cortical surfaces, and as such, adapts the filtering process to the underlying morphological details at each point in the cortex.

Face shape classification is considered one of the trending topics in the artificial intelligence research field. Face shape classification can be employed in many broad-scoped projects, such as hairstyle recommendation systems in the beauty and fashion industry. In this paper, the inception v3 model was employed to reach the highest possible performance for classifying the different face shapes. The model was re-trained after applying a proposed sequence of preprocessing techniques, including image straightening, cropping, resizing, and normalization. The model was re-trained on different data sizes of females' images. Comparing the proposed model to the other state-of-art previous work showed an outperforming testing accuracy of 94.3%.

A method based on Hidden Markov Model (HMM) and Mel Cepstral Coefficients (MFCC) as the extraction feature parameters is proposed to realize footstep recognition. The footsteps are collected and preprocessed. The preprocessing adopts the method of endpoint detection to determine the start and end of the sound signal. Then the feature parameters MFCC are extracted, and the footstep feature library is trained based on the hidden Markov model. Input the footstep sound test sample, and use the Viterbi algorithm to perform feature matching with the trained footstep sound feature library to identify the sample identity. The experimental results show that the method has a higher recognition rate.

Based on the characteristics of electroluminescent circuit board, a new automatic optical detection technique is proposed. A visual detection system for circuit board fault is designed, which combines digital image processing and sensor technology. Image preprocessing, image registration, luminescent region extraction and fault location are carried out. Different detection methods are proposed for different circuit boards. The system can realize the intelligent fault detection of the circuit board, solve some problems in the traditional manual detection, meet the real-time and high precision requirements of the fault detection of the circuit board, and is easy to be implemented in engineering.

Electronic invoice has been gradually popularized in financial reimbursement. The automatic recognition of electronic invoice image can help financial personnel save processing time and reduce entry errors, so it is great significance. This paper provides a method for recognition of electronic invoice image, which mainly includes invoice preprocessing, invoice positioning, invoice cutting, sample training and building of convolution neural network model. Experimental results show that this method has high recognition accuracy and can effectively help financial staff reduce their workload.

Every year, doctors diagnose skin cancer in around 3 million or more patients across the globe. Currently, it is one of the most widely recognized kinds of cancers for human health. Hence, we need an early diagnosis to prevail any critical condition of the infected patients. Apparently, it can treat with topical drugs, if it diagnoses in an early stage. Hence as an outcome, skin cancer is responsible for less than 1% of all cancer deaths. There are two types of tumors in the skin cancer diseases domain, such as benign and malignant. To develop a robust and early screening system to diagnose skin cancer, it requires an efficient algorithm for prediction, trained with a large dataset. The primary aim of this research is to develop an efficient skin cancer screening process using a robust deep neural network with a large dataset. In this paper, we intend to determine considerate and dangerous types of skin cancer tumors using dermoscopic images from a publicly available dataset. We proposed an efficient and fast scaled 2D-CNN based on EfficientNet-B7 deep neural architecture with image preprocessing. This paper also uses two different pre-trained deep neural architectures, such as VGG19, and ResNet-50 to compare the performance with the proposed architecture. The proposed architecture outperformed the other pre-trained CNN models whereas the proposed architecture achieved higher AUC and accuracy compared to other architectures.

The aim of this paper is to present some results on word embeddings for the Romanian language, based on the word2vec method. More concretely, we generate word embeddings of different lengths, and using different preprocessing and training techniques. The embeddings are general purpose, and we use the Romanian language version of Wikipedia as corpus. We also evaluate the computational resources needed for the task. The embeddings are validated by performing some experiments on synonyms detection, using a new dataset created for this purpose. The code and the dataset are made publicly available. The results indicate that these types of embeddings can be used with the summarization approaches.

This paper proposes a tool for studying the relationship of surface electromyography between healthy people and someone with some pathology in the lower limb. Support Vector Machine SVM is used to classify electro myographic signals because models are robust to overfitting. For the lower limb, analysis has been taken the EMG Dataset from UCI Machine Learning [1]. This database contains signals from 11 subj ects with knee abnormality and 11 normally, previously diagnosed by a professional. They undergo three movements to analyze the behavior associated with the lower limb, gait, leg extension from a sitting position, and flexion of the leg up. Analysis was divided into 4 stages: preprocessing, features extraction, training, and validation. Several conventional electromyography features are used in performance comparison with it combined with Enhanced features. Based on results obtained by algorithms with different Kernel the Support Vector Machine model provided by MATLAB® Classification Learner app achieves the highest accuracy of 96.7%.

The focus of this work is to test the security offered by Text-based CAPTCHAs. We present different types of CAPTCHAs and a preprocessing and segmentation process to clean noise in CAPTCHA images and crop digits or characters in single images. We present a convolutional neural network architecture trained under several hyperparameters, comparing multiple models with different batch sizes, epochs, and optimizers. We confirmed that using Text-based CAPTCHAs is no longer a secure mechanism for protection because, with simple computer vision techniques and current machine learning algorithms, they can be broken. We achieved a 90.49% accuracy with our model trained with a mix of four datasets and up to 97.10% with one dataset, which is enough to consider these schemes insecure in practice.

Systems used for pattern recognition are usually divided into 4 stages: signal acquisition, preprocessing, feature extraction and classification. However, the use of algorithms in these last 3 stages is not justified and researchers use them without criteria other than the result achieved at the end of the process. In this paper we propose a new methodology and show its particular application to the recognition of five hand gestures based on 8 channels of Electromyography using the Myo armband device placed on the forearm. If n features are extracted, they will form clusters of points in n-dimensional space and now the selection of the best preprocessing algorithms and the best features are based on maximizing the distance among clusters. On the other hand, both intrapersonal and interpersonal variability are treated to facilitate the understanding of the phenomenon. As a demonstration, it was applied to 12 people and the recognition accuracy was 97%.

The most prevailing and most deadly form of the brain tumor is glioma, with a mortality rate of around 70%. Therefore, early and efficient treatment is necessary for improving patients' wellbeing. The MRI method has become a popular method for identifying these tumors, but the amount of data retrieved limits the use of quantitative measures that are accurate and reproducible. Therefore, Automating brain tumor segmentation from MR images is essential for tumor analysis and monitoring. For successful tumor delineation, accurate and reliable tumor segmentation techniques are used for gliomas, which are aggressive and heterogeneous. With the use of CNNs, which are widely utilized for biomedical image segmentation, the accuracy of the current state-of-the-art has greatly improved on the problem of segmenting brain tumors. This paper proposes a M U-Net architecture with a PReLU activation layer and smaller input kernels to increase the performance of segmenting tumor regions in MR images. An augmentation of data combined with normalization during preprocessing is applied to rectify the issue of class-imbalance in tumor segmentation. This experiment uses the BRATS 2017 dataset. It achieved accuracy, sensitivity, specificity of 97.94%, 99.9%, and 92.58% respectively.

The research of clinical image processing gathers its essentialities with its growth in the requirement of efficient and automated diagnosis in minimal duration of time. Apart from that the retrieval system of clinical images has to be provided as the tool for clinicians and radiologists for acquiring the clinical images as that of the fault image present in the content. Magnetic Resonance Imaging (MRI) is an emerging technique in the field of medical image acquisition and processing that has played an essential and vital role in the research field of neuro science in order to study the images of brain. Classifying the brain images to different clinical state is very important task for clinicians to make decisions for proceeding with the treatment to cure major deadly disorders. Hence it is considered as an essential sector in the retrieval system for distinguishing the normal subjects from the abnormal one. Brain tumor is considered as one of the most important conditions of abnormalities which results in death at most of the cases. In this proposed study, the brain images are taken as a cluster of databases with different clinical conditions in both normal and abnormal state with which 80% of data are utilized for training the classifier and remaining 20% of data are being used as test data. The sequence of process that are used for detecting the clinical state of MRI images are followed by preprocessing, segmentation, feature extraction, feature selection and classification. The multi kernel Support Vector Machine using Radial Basis Function is being used for classifying the MRI images. The performance metrics such as sensitivity, specificity and accuracy are being determined and it is realized to be as an efficient strategy that is useful for categorizing the clinical state of brain whether it is a normal one or abnormal one. It is being inferred that the classification accuracy is being attained with 98.62% in predicting the clinical normality and abnormality conditions.

The Electrocardiogram (ECG) is a bioelectric record of heart activity. ECG signal classification is crucial for heart disease diagnosis. It is difficult to make an appropriate ECG classification. In this study, MIT-BIH Arrhythmia ECG data base from physionet has been used to classify Cardiogram Signals. The very first, raw ECG signal is preprocessed by using noise cancellation (filtering) methods are Adaptive noise cancellation techniques, low pass linear noise cancellation techniques and high pass linear filter. After preprocessing reduces the dimensionality of data by using Coordinated Factor Analysis (CFA) and finally, dimensionally reduced signals are classified by using Water Wave Optimization (WWO) classifier. The Performance metrics like Performance Index, Accuracy, Sensitivity, and Specificity are evaluated using WWO to analyze the classification of arrhythmias from ECG signals.

Guess of Discrete Cosine Transform (DCT) is valuable for diminishing its computational intricacy without critical effect on its coding execution. This paper presents a summed up recursive calculation to get symmetrical approximation of DCTs at the expense of augmentations for input preprocessing. The recursive scanty framework decay is performed and it utilizes the balances of DCT premise vectors for inferring the estimated calculation. It is exceptionally versatile for Simulation based equipments just as programming execution of DCT of higher lengths and it can utilize the current estimate of 8-direct DCT towards getting rough DCT of any force of length two. A completely adaptable reconfigurable equal engineering for the calculation of inexact DCT is planned dependent on the proposed calculation. The proposed reenactment dependent on design is found to offer numerous benefits as far as equipment intricacy, and versatility. The equipment apparatus utilized are double center processor and FPGA Spartan 3 EDK pack and the tool compartment is MATLAB 7.8.

Biometric system has gained more importance in providing high security in individual identification as it uses a network of blood vessels underneath the palm skin. This paper proposes a new algorithm for palm vein identification using a histogram of gradient t(HOG). Raw images of palm hand vein are taken from a public dataset named VP base dataset. Region of interest was extracted after preprocessing stage, after that essential features were extracted by following different steps of the HOG algorithm, which captures edge information inside the images, where they are the vital features containing valuable information. For the purpose of classification Support Vector Machine was used. By experimental work, our algorithm gives promising results than different existing feature descriptors.

Epilepsy is a central nervous system disease that affects loss of consciousness and convulsions. The testing of electrical activity in the brain is identified as EEG. Ability to predict the onset of epileptic seizures before their commence seems to be very useful for medication-assisted seizure protection. Electroencephalogram (EEG) signals are now used to predict epileptic seizures using machine learning techniques and statistical analysis. Pre- processing of EEG signals for dynamic range and extra functionality, but in the other hand, extraction are two real issues that have a negative effect on the system’s run time and correctness. Therefore, this paper proposes a model that provides a reliable method for both pre-processing and classification. It is implemented using FIR filter for preprocessing and machine learning classification methods. Among the variety of machine learning algorithms, this paper uses RF algorithm and BPNN for classification. After the classification, the results were compared, and it is revealed that Random Forest classifier gives 95% accuracy which is more efficient than Neural network via back propagation.