Search

For the detection of product color uniformity and seam texture defects in the production process of test paper, detection methods are proposed respectively. For the detection of the color uniformity, a uniformity detection method based on dynamic light correction is proposed. The illumination correction image is used for color dynamic compensation of the test paper image to solve the influence of uneven illumination on the test paper. For the detection of texture defects such as seams, a texture defect detection method based on gray gradient aggregation intensity is proposed. This method uses one-way Gaussian convolution to strengthen features, and then calculates the gradient, divides the region and calculates the gradient extremum aggregation strength to judge texture defects, which can quickly and efficiently find seam defects. The experimental results show that the above methods have obvious effects on the color uniformity and seam detection of the test paper. The color uniformity accuracy rate is 96%, the seam detection accuracy rate is 97%. The total detection time of a single test paper is about 90ms. The algorithm is real-time and accurate, and can be used in actual production detection.

Paper dating analysis is an important research direction of document inspection, which is widely used in the inspection and identification of cultural relics and ancient books. This article focuses on the analysis of paper dating in ancient books, and proposes a non-destructive inspection and analysis method based on paper fiber texture images. Aiming at the global stacking morphology of paper fibers and the local features of specific types of fiber morphology, we propose a neural network-based hybrid texture feature extraction and representation method: On the one hand, we use convolutional networks to obtain global features of fiber texture; at the same time, we designed a method of extracting and representing local fiber morphological features based on the attention mechanism. By mixing the above two types of features, we realize the extraction and representation of paper fiber features. Furthermore, we use the GRU(Gate Recurrent Unit) model to establish a paper dating time series model and design a new loss function. In order to verify the method, this paper selects 36 domestic books published from 1950 to 2000, and uses the document checker VSC 6000 to collect paper texture images as a dataset, and verifies the effectiveness of the proposed method on this dataset. Experiments prove that the method proposed in this paper has achieved ideal results in paper dating analysis.

In the last several years, we published several papers to address the problem of Face Identification. The techniques employed in those articles were implemented in transform domains. The Discrete Cosine (DCT) and the Discrete Wavelet (DWT) Transforms were utilized, either combined or individually, to extract features which form the final model for each participant in a given dataset. In this paper, we highlight significant parts of our previous works in order to give a fair comparison among all approaches. The results included here are for the following datasets: ORL, YALE, FERET, FEI, Georgia Tech, and Cropped AR. Features are DWT, DCT, energy-based selected DCT-DWT, and combined DCT-DWT coefficients while the classifier is Euclidean distance, either squared or with power of one.

Visions have an important role in guiding and legitimizing technical research, as well as contributing to expectations of the general public towards technologies. In this paper we analyze technical robotics papers published between 1998 and 2019 to identify themes, trends and issues with the visions and values promoted by robotics research. In particular, we identify the themes of robotics visions and implicitly normative visions; and we quantify the relative presence of a variety of values and applications within technical papers. We conclude with a discussion of the language of robotics visions, marginalized visions and values, and possible paths forward for the robotics community to better align practice with societal interest. We also discuss implications and future work suggestions for Responsible Robotics and HRI research.

Even current OCR technology has made significant progress with the help of deep learning, rapid identification algorithms for specific information in documents still need elaborate design. In this paper, a lightweight operating framework based on deep neural network is established to deal with the extraction of date and time information from paper worksheets used in transformer station. In the framework, YOLOv4-Tiny is employed to detect the locations of the dates and the time in document images which are mixed with printing and handwritten fonts. Since the handwritten dates in pictures are difficult to segment, a novel lightweight CNN named MobileDateNet is proposed that can recognize the date serially without segmentation. The experimental results show that the identification accuracy of the whole framework reaches 92.1% at the average running speed of 36.5 FPS on an NVIDIA 3090 GPU, ensuring the recognition accuracy and the operating efficiency.

Forgery of paper invoices and certificates negatively causes huge economic loss every year. In the molding process of paper, plant fibres inside will distribute randomly and form the unique microscopic surface. Accordingly, forensic researchers have presented many techniques to build its feature descriptor for paper anti-counterfeiting. Among those methods, normal vector field and reconstructed surface based on multi-view photos captured by mobile camera have indicated incredibly effectiveness and practicality. However, the inefficiency of existing models using mobile camera is the neck in real time detection scenarios. In this paper, we focus on the efficient microstructure orientation estimation of paper surface for authentication. To address this problem, we investigate the reflection characteristics of rough surface and quantify the light attenuation accurately. Then, we model the entire propagation path of light emitted by flash, and exploit single photo captured by mobile camera to reconstruct microscopic inclination based on beam energy loss. Reduction of required pictures significantly enhances forensics efficiency. Through comparison experiments, we demonstrate that our proposed method can be used for paper discrimination effectively with speedy feature extraction.

The experiment of quality assurance of polymeric outdoor insulators service life is necessary to be checked the class of wettability by using the digital image processing tools. The image processing can improve its pictorial information for human interpretation, render it more suitable for autonomous machine perception. This paper provides a survey of the recent technology and theoretical concept to explain the development of the Hydrophobicity Class (HC) of porcelain insulator. The hydrophobic properties of the housing material were found degraded to a different extent between field-aged insulators due to differences in material structure and pollution conditions. This paper contributes to the latest development of reviews relating to image processing for the implementation of porcelain insulator hydrophobic systems testing., and their related studies. We categorized the computer vision mainstream into 2 group such as Machine Learning Methods Applied to Digital Image Processing and Deep Learning Methods Applied to Digital Image Processing. Then compare pros/cons in each type of technique. We also provide brief explanation on the up-to-date information about the techniques and their performance.

In modern academic social network, it is very difficult for scholars to find academic papers consistent with their research direction. Time is a critical factor in paper recommendation. As time goes on, the impact of an academic paper would gradually fade. Likewise, the research interests of users may also change. Therefore, we propose a temporal perceptual neural collaborative filtering model that integrates the multi-dimensional features of papers. We conducted our experiments on the dataset from CiteULike, comparing the recommended results by using four time-decay functions and evaluating our model with multiple evaluation indicators. The satisfactory results show that our model is effective in filtering out the expired papers by considering the characteristics of papers and the changes of scholars' interests.

In this paper, we propose a data augmentation framework for Optical Character Recognition (OCR). The proposed framework is able to synthesize new viewing angles and illumination scenarios, effectively enriching any available OCR dataset. Its modular structure allows to be modified to match individual user requirements. The framework enables to comfortably scale the enlargement factor of the available dataset. Furthermore, the proposed method is not restricted to single frame OCR but can also be applied to video OCR. We demonstrate the performance of our framework by augmenting a 15% subset of the common Brno Mobile OCR dataset. Our proposed framework is capable of leveraging the performance of OCR applications especially for small datasets. Applying the proposed method, improvements of up to 2.79 percentage points in terms of Character Error Rate (CER), and up to 7.88 percentage points in terms of Word Error Rate (WER) are achieved on the subset. Especially the recognition of challenging text lines can be improved. The CER may be decreased by up to 14.92 percentage points and the WER by up to 18.19 percentage points for this class. Moreover, we are able to achieve smaller error rates when training on the 15% subset augmented with the proposed method than on the original nonaugmented full dataset.

Lately, Deep Learning has indicated its capacity by adequately tackling complex learning issues which were not possible previously. Specifically, Convolutional Neural Networks are most generally utilized and have indicated their viability in image detection and recognition problems. Convolutional Neural Networks(CNN) can be used to perform different kinds of tasks. However, to recognize even a single image, billions of calculations are required. For example, ResNet 50 takes 8 billion calculations just to figure out what's in a single image. So we need to find out ways to be able to run that superfast and that's why we need hardware accelerators. In this paper the basic information about Convolutional Neural Networks is provided along with the key operations involved and the brief idea about Field Programmable Gate Arrays(FPGA) is given which enable them to be used for accelerating the inference process of Convolutional Neural Networks. Various techniques which were employed previously for accelerating the Convolutional Neural Networks are discussed. We are focussing on designing a energy efficient Field Programmable Gate Arrays accelerator for accelerating the inference process of Convolutional Neural Networks with reduced latency and increased throughput.

People re-identification is a fundamental stage in video surveillance systems. This stage is used for determining the labels of the people’s images usually based on their visual similarities. The appearance salience caused by wearing the clothes and carrying objects, with specific colors can be used for improving the performance of re-identification systems. To address the mentioned point, in this paper, we propose a color-based re-ranking process. In the proposed approach first, the input images are semantically segmented as: person’s body, possible carried objects, and background. Then the most frequent colors of person’s body and possible carried objects are used to re-order the ranked results obtained from two commonly used re-identification approaches i.e., Gaussian of Gaussian (GOG) and Hierarchical Gaussian Descriptors (HGDs). Experimental results on various people re-identification datasets show that our proposed re-ranking approach can improve the performance of re-identification systems.

Due to the increasingly growing demand for user identification on cell phones, PCs, laptops, and so on, face anti-spoofing has risen to significance and is an active research area in academia and industry. The detection of the real face then recognize it present an important challenge regarding the techniques that can be used to spoof any recognition system like masks, printed photos. This paper we present an anti-spoofing face method to solve the real-world scenario that learns the target domain classifier based on samples used for training in a particular source domain. Specifically, with the conventional regression CNN, the Spatial/Channel-wise Attention Modules were introduced. Two modules, namely the Spatial-wise Attention Module and the Channel-wise Attention Module, were used at spatial and channel levels to improve local features and ignore the irrelevant features. Extensive experiments on current collections with benchmarks datasets verifies that the recommended solution will significantly benefit from the two modules and better generalization capability by providing significantly improved results in anti-spoofing.

Drowsiness can be dangerous when performing tasks that require constant attention, such as driving a vehicle. Sleepiness is correlated with a variety of physiological variables, such as eye closing, head movements, pulse rate, eye twitch rate, etc. Also, the yawn can be considered as an accurate indicator of drowsiness and fatigue. Yawning detection is very important for the safety purpose of drivers as it will let the driver know if he/she is getting drowsy. Driving at that moment may not be safe. Several automatic yawning detection techniques have been developed for driver's drowsiness monitoring system. Nevertheless, correctly detecting the yawning of the driver and predicting exhaustion in real-time situations is still a crucial challenge. In this paper, we will be reviewing various existing machine learning approaches for driver's yawning detection. In previous approaches, various classical machine learning algorithms such as viola-Jones, contour activation algorithm and SVM have been used for yawning detection, but these approaches failed to predict yawning in realtime situations. Using Deep learning techniques, we can make a real-time yawn detection system with high accuracy. We find that some precious Deep learning algorithms like CNN, RNN, LSTM, Bi-LSTM can detect the patterns with high accuracy. After the comparison of various algorithms and techniques, we find that with the help of Deep learning algorithms the yawning can be detected in real time with high accuracy.

Human recogntion technologies are gaining significant research attention, where the model can be trained to be more precise to recognize the poses performed by objects. Activity identification is a kind of problem, which needs more research consideration and improvement. There is also an increasing need to detect different poses of objects. To tackle the issue, different sensors like Gyroscope and Accelerometer are required to classify the data in the form of images using machine learning algorithms like SVM and CNN. These approaches help us in implementing real-time applicaions such as health monitoring, tracking sports activity and security. The paper also discusses about its benefits, limitations and prominent approach for human activity recognition.

This paper is review on image fusion for different technique i.e. Non-subsampled Contourlet Transform (NSCT) and Neural Network (NN). NSCT is the pyramid structure and gives the multiscale feature of the image. NSCT used to directional filtering structure and used to filter bank. Neural network is a progression of calculations that tries to perceive fundamental connections in a bunch of information through an interaction that impersonates the manner in which the human mind works. The implemented algorithm will applied to MRI image and with the help of NN to calculated accuracy. The previous algorithm was work of different technique i.e. principal component analysis, discrete wavelet transorm and curvelet transform. The implemented algorithm will implement MATLAB software and calculate mean square error and peak signal to noise ratio.

This paper proposes an intelligent cockpit based on artificial intelligence technologies such as voice recognition ASR, TTS, NLP, and multi-round dialogue. The cockpit monitors the progress of the distribution network virtual dispatcher and analyzes the decision engine, and replays the virtual dispatcher online. The implementation process and thinking progress of the virtual dispatcher are overall monitored, and the online monitoring and safety monitoring of the operating personnel are supported.

In order to automate the task of correcting handwritten subjective answers, we envision a system that reads such answers and grades them. In this project, we have created a mobile application to achieve this task. We have used the RAKE algorithm to extract key-phrases, which are then embedded using Word2vec model and compared using Word Mover's Distance. Our created model achieved almost human like evaluation accuracy. We have also provided accessibility features to the users with the aim of making this application a one-stop solution to automatic evaluation of handwritten subjective answers.

Sentiment analysis and opinion mining have developed significantly throughout the last decades. This study aims to assess, amongst many other aspects, people's feelings, opinions, and emotions of someone or something. This review's principal objective is to automatically assess the direction of the review and contrast this with the evaluation made by the reviewers. This will allow scientists to categorize and evaluate reviews cross-sectionally and encourage a scientific article's overall evaluation more critically. An EM-VC ensemble approach has been proposed for analyzing the reviews. This approach uses POS to achieve a syntactic sentence structure. With dictionaries, this syntactic structure enables the semantic orientation of the analysis via the scoring algorithm to be determined. To test the proposed approach's efficiency and performance about the baseline method, some tests have been performed using typical performance metrics like precision, accuracy, F1-score, and recall. The findings reveal that improvements are achieved in binary classification, ternary classification, & five-point classification compared to the existing HS-SVM method. Still, it provides a slight improvement in accuracy in the context of multiclass classification compared to others.

In my talk I will discuss the AI and brain-computer interface (BCI) technologies overview in the context of integration with robotic and biological systems starting from classical Rosenblatt perceptron till current breaking through technologies of BCI that could change the scientific outlook of the field. This is what we currently broadly call now AI and neurotechnology. Two bright examples are ResNet inspired by the projections of a mammalian cortical column and U-Net with close to ResNet ideas and topology. The other interesting cortical column inspired NN architecture is hierarchical temporal memory introduced by the Numenta company. These brain or neuro inspired approaches are widely used in machine learning, computer vision, natural language processing, path planning and broader in robotics.In the neurotechnology field I want to reference a lot of interesting and projects dedicated to BCI. Starting from works of Miguel Nicolelis where researchers connected the robotic hand via a computer system with a motor cortex of a monkey and later Lebedev group demonstrated the adaptation of a mammalian brain to create the representation of robotic limb extending biological limbs. I want to pay the special attention to works of Kevin Warwick that implemented invasive nervous system to nervous system interface and the robot managed by neurons of a rat brain developed and trained during the experiment.The commercial company Neuralink recently demonstrated the most advanced 1536-channel BCI exploiting wireless Blue-tooth interface to record and process a brain activity. Elon Musk in his Neuralink presentation last year demanded that one of the nearest goals of the company is spinal cord injury and the recovery of the lost motor control of limbs. Several projects in neurorehabilitation targeted to the motor control demonstrated recently important success using medical programmed neurostimulators.The perspective but currently in early state of the development is the direction of neurosimulation based models that could be used for the robotic system control (exoskeleton) or as neuroprosthesis as the part of closed loop system for example projects simulating the spinal cord, where I can see interesting future opportunities both in medicine and robotics.

The aim of this work is to classify the dementia stage using microstructural white matter (WM) diffusion kurtosis indices. We develop various class imbalance machine learning methods in classifying the cognitive normals (CN), Alzheimer's disease (AD) and mild cognitive impairment (MCI) individuals. Diffusion-weighted images of 155 subjects were collected with 44 CN, 88 MCI and 23 AD individuals aged between 60 to 96 years. We first perform skull striping using FSL tool along with head motion and artifact correction. We calculate the mean diffusivity, fractional anisotropy from various brain regions of WM indices from diffusion kurtosis imaging (DKI) using tract based spatial statistics. Then we estimate the kurtosis fractional anisotropy of various WM regions. After estimation of all the regional KFA indices, we trained these indices using various class imbalance machine learning classification algorithms, as the sample size is different in all the three groups. We found that balanced random forest classifier was the best classifier with an accuracy of 74% when compared to other methods for 5-fold cross-validation. We conclude that class imbalance machine learning methods are potential in classifying cognitive normals, AD and cognitive impairment individuals.

In autonomous driving, 3D object detection is es-sential as it provides basic knowledge about the environment. However, as deep learning based 3D detection methods are usually computation intensive, it is challenging to support realtime 3D object detection on edge-computing devices in selfdriving cars with limited computation and memory resources. To facilitate this, we propose a compiler-aware pruning search framework, to achieve real-time inference of 3D object detection on the resource-limited mobile devices. Specifically, a generator is applied to sample better pruning proposals in the search space based on current proposals with their performance, and an evaluator is adopted to evaluate the sampled pruning proposal performance. To accelerate the search, the evaluator employs Bayesian optimization with an ensemble of neural predictors. We demonstrate in experiments that for the first time, the pruning search framework can achieve real-time 3D object detection on mobile (Samsung Galaxy S20 phone) with state-of-the-art detection performance.

Graph neural networks (GNN) have achieved state-of-the-art performance on various industrial tasks. However, the poor efficiency of GNN inference and frequent Out-of-Memory (OOM) problem limit the successful application of GNN on edge computing platforms. To tackle these problems, a feature decomposition approach is proposed for memory efficiency optimization of GNN inference. The proposed approach could achieve outstanding optimization on various GNN models, covering a wide range of datasets, which speeds up the inference by up to 3×. Furthermore, the proposed feature decomposition could significantly reduce the peak memory usage (up to 5× in memory efficiency improvement) and mitigate OOM problems during GNN inference.

It is difficult to evaluate a person’s ability or performances because there is no common yardstick for measuring individual human abilities. For example, in general, paper based tests have different characteristics (difficulties, accuracies, and so on) because they depend on items of each test. Recently, e-testing technologies from AI approach, for which each test form has equivalent measurement accuracy but with a different set of items, have become popular. Even if different examinees with the same ability take different tests, their test scores should be guaranteed to be equivalent. We introduce state-of the-art e-testing technologies from AI approach.

the real datasets in the world usually are imbalanced; the number of samples for their classes is not equal. Classifying these datasets makes the classifiers pay attention to the class with more samples than the classes with fewer samples. The Qur’anic dataset can be considered an imbalanced dataset because verses of the Qur’anic topics are not equal. Many studies have been performed to classify Qur’anic text using different classifiers. However, few studies classified the Qur’anic verses based on Imbalanced Learning (IL). So, this work aims to classify the Qur’anic text using Ensemble methods, Boosting and Bagging. The base classifiers of these methods were LibSVM, Naive Bayes, KNN, and J48. Three techniques are conducted in this paper based on the standard classifiers. The three techniques are: implementing the base classifiers alone, implementing these classifiers with the Boosting method, and implementing the classifiers with the Bagging method. The results showed that the Quranic classification performance was improved when the ensemble methods were applied for the imbalanced Qur’anic verses in the standard classifiers.

Image classification needs to consider image degradations in practice because an image classification network trained with clean images works poorly for degraded images while digital images usually include some degradations such as JPEG compression. To tackle this problem, a common approach is training classification networks on degraded images with various levels of degradation, e.g. various quality factors for JPEG compression. However, the classification networks do not usually have enough accuracy for clean images because the classification networks have been averagely trained for various levels of degradation. This paper aims to construct a classification network of degraded images without sacrificing the performance of clean images. This paper proposes a network to learn the classification of degraded images and degradation levels of degraded images as multi-task learning. Learning the classification of degraded images is the main task of multi-task learning. On the other hand, learning degradation levels is a sub-task of multi-task learning and reinforces the classification ability. In the proposed network, a feature extractor of the classification is trained with image features acquired from a classification network of clean images by using consistency regularization with the cosine similarity loss. The experimental results using different types of degradations, including JPEG compression, Gaussian blur, Gaussian noise, and salt-and-pepper noise, show that our proposed network has enough ability to classify degraded images without sacrificing the performance for clean images.

With the development of personal broadcasting such as YouTube, the demand for editing the video he filmed himself is steadily increasing. Traditional media video editing solutions help edit the results of the filming. However, there is a problem that requires a lot of editing time, as people usually edit videos that are from tens to hundreds of times longer than the final result of a personal broadcasting after filming. To overcome this problem, this paper automatically classifies images with specific scenes in the whole media image editing process, and secondly proposes automatic media editing solution technology in which people intervene. In particular, personal broadcasting focuses on the use of images that include characters, specific objects, and cue sign gestures among the entire. While the existing deep learning techniques such as faces, objects and gestures are advanced, integrated recognition technologies that simultaneously deal with special requirements for editing videos are still in the early stages of research. In this paper, the automatic composite recognition technology for editing video based on deep learning is proposed. The proposed technology was implemented with python and tensorflow software based on edge computing equipment. Using actual youtube videos, it took 0.1 second to process five-person recognition, 63-food recognition, or cue sign recognition using clapping or V poses at the same time. The recognized results are divided into timestamps of the entire movie, recognition results, and locations of objects on the screen, and are output to the json file. In addition, this solution was developed on an edge computing in order to increase real-time reliability. We expect to provide automatic video editing based on perceived json results as well as shorter editing times based on this implementation.

Since its inception in 2015, Style Transfer has focused on texturing a content image using an art exemplar. Recently, the geometric changes that artists make have been acknowledged as an important component of style[42], [55], [62], [63]. Our contribution is to propose a neural network that, uniquely, learns a mapping from a 4D array of inter-feature distances to a non-parametric 2D warp field. The system is generic in not being limited by semantic class, a single learned model will suffice; all examples in this paper are output from one model.Our approach combines the benefits of the high speed of Liu et al. [42] with the non-parametric warping of Kim et al. [55]. Furthermore, our system extends the normal NST paradigm: although it can be used with a single exemplar, we also allow two style exemplars: one for texture and another geometry. This supports far greater flexibility in use cases than single exemplars can provide.

Automotive radar sensors offer a promising and effective modality for perception and assessment of the surrounding environment. A key element of environment sensing in automotive radars is the reliable detection, classification and tracking of vulnerable road users such as pedestrians and cyclists. In this paper, we propose an integrated Bayesian framework dubbed BayesRadar, which improves the overall classification accuracy by tracking the embedding vector of a neural network and its prediction uncertainty via recursive Kalman filtering over time. Apart from the classification accuracy of a model, a critical measure includes the analysis of statistical confidence over the target class score. Such measures for predicting the true correctness likelihood of the classification estimates are essential in safety-critical automotive applications. Therefore, in this paper, we present and evaluate the classification, embedding cluster score and statistical confidence performance of the proposed framework in the context of classifying vulnerable road users compared to state-of-art deep learning approaches. Furthermore, we demonstrate superior performance of the BayesRadar for unseen classes compared to long short-term memory based temporal tracking of the embedding vectors.

Machine learning is presently acknowledged as a significant ingredient of research in many fields, including robotics. The use of robots to perform assorted tasks is evident in difficult, uncompromising, and hazardous spaces and sectors such as manufacturing, transportation, healthcare, landmines, mining, patrolling, disaster relief etc. For a robot to carry out its assigned task, it normally has to navigate safely without collisions to different locations, which also means understanding its working environment, collectively known as the robot navigation problem. This paper considers finding a solution using neural networks to the robot navigation problem, particularly the path planning problem that includes fixed obstacles. The objective of the path planning problem is to find a route to the final destination that is optimal and also collision-free. Different training algorithms and network structures are used to construct models that can predict a turning angle for the point-mass robot which will be used to avoid obstacles in the robot's path to the destination. This paper will present a comparative analysis of the performance of different feedforward neural network models. The results suggest that the feedforward neural network model with 10 neurons and Bayesian regularization performed the best. The model has been used to avoid obstacles in two different environments. The trajectories show that the robot has safely avoided obstacles in its path and reached the destination.

Sign languages is a critical requirement that helps deaf people to express their needs, feelings and emotions using a variety of hand gestures throughout their daily life. This language had evolved in parallel with spoken languages, however, it do not resemble its counterparts in the same way. Moreover, it is as complex as any other spoken language, as each sign language embodies hundreds of signs, that differs from the next by slight changes in hand shape, position, motion direction, face and body parts contributing to each sign. Unfortunately, sign languages are not globally standardized, where the language differs between countries and has its own vocabulary and varies although they might look similar. Furthermore, publicly available datasets are limited in quality and most of the available translation services are expensive, due to the required skilled human personnel. This paper proposes a deep learning approach for sign language detection that is finely tailored for the Egyptian sign language (special case of the generic sign language). The model is built to harnesses the power of convolutional and recurrent networks by integrating them together to better recognize the sign language spatio-temporal data-feed. In addition, the paper proposes the first Egyptian sign language dataset for emotion words and pronouns. The experimental results demonstrated the proposed approach promising results on the introduced dataset using combined CNN with RNN models.

© 2021 IEEE.Coronavirus usually transmits from the animal to the human, but now, the virus transmission is between persons. Therefore, scientists and researchers are trying to develop several types of machine learning methods to defend against COVID-19. Medical images play a significant role in this time due to they can be used to recognize COVID-19 accurately. However, in this paper, we used X-Ray images, the images undergone to sharpening techniques to increase the results further. The texture techniques named local binary pattern (LBP) have been used in order to extract features. The features obtained were applied to the support vector machine (SVM). The results we achieved were 100% for all performance measurements. Our results were conspicuously superior compared to the state-of-the-art papers published.

In this paper three different types of architectures have been trained and tested to determine which one of them is more efficient in detecting exudates from retinal images. The database that was used is SUSTech_SYSU from which 100 images were extracted for image testing and another 100 images that were used in obtaining the training dataset. The three architectures that were studied in this paper are SqueezeNet, ResNet50 and GoogLeNet. The results showed that from the three architectures the most efficient in exudates detection was ResNet50 with an average accuracy of 0.928, followed by SqueezeNet with 0.910 and finally, GoogLeNet for which the average accuracy was 0.907.

This paper presents a method for radar-based contactless vital sensing and ECG (ElectroCardioGram) signal reconstruction using deep learning. ECG signal is a typical representation of heartbeat signals. However, its measurement usually requires any contact device, which is not suitable due to discomfort of subjects. Radar system is effective for vital sensing because it can measure small displacement of body surface caused by breathing and heartbeat without contact. On the other hand, most of the methods using radar system are limited to evaluating simple indices such as heart rate and heartbeat interval while subjects or devices are stationary. In this paper, we propose a method for body surface displacement signals using FMCW (Frequency-Modulated Continuous Wave) radar and reconstructing ECG signals using CNN (Convolutional Neural Network). The result of experiments on six healthy males shows the ECG signals are successfully reconstructed. Furthermore, we confirmed that the proposed method can reconstruct signal waveforms even in an environment with low SNR (Signal-to-Noise Ratio).

Power electronic devices such as Insulated Gate Bipolar Transistors (IGBT) are widely used in electronic systems. The failure of IGBT may bring fatal damage to the whole system. The ability to accurately predict the remaining useful life (RUL) of IGBT can provide a basis for intelligent equipment maintenance, thereby minimizing the risk of catastrophe failure at work. The research in this field aims to achieve accurate and timely RUL prediction of equipment. This paper exploited the IGBT accelerated aging data from NASA Ames Research Center, and the monitoring data of IGBT is collected under thermal overstress load condition with square signal bias at gate. This paper developed three machine learning models to predict the RUL of IGBT, including Back Propagation Neural Network(BPNN), Random Forest and Extreme Learning Machine. These models realize real-time prediction of RUL, and the performance of these models is observed and analyzed.

Rapid and accurate detection of garbage has become a hot issue in the field of computer vision. In this paper, we propose an intelligent garbage detection algorithm to improve the accuracy and speed of garbage detection. For the self-made garbage dataset, to reduce the annotation workload and the problem of missing annotation and wrong annotation, YOLOv4 combined with self-learning strategy is used to realize the training of garbage detection model and improve the accuracy of detection tasks. At the same time, the fast NMS algorithm is introduced to accelerate the prediction process and improve the model prediction speed. The experimental results show that the detection effect of the algorithm in this paper is significant, and it can effectively identify six different kind of garbage, with mean Average Precision (mAP) of 95.36% and a detection time of 25ms for a single image, which is expected to be applied to the recognition and detection of waste targets in industrial conveying scenarios.

Accurate land use information extraction plays an important role in land management, environmental monitoring, urban and rural planning and development and ecological assessment. High resolution remote sensing images have the characteristics of class similarity. Tradi¬tional methods of extracting ground object information from remote sensing images are limited to specific scenes and data sources, and have low segmentation accuracy and efficiency. In this paper, we shed the light to this research topic. To further improve the accuracy of land use information extraction, this paper takes HRNet(High-level Resolution Network) as the backbone network, which improves the performance of the network by fusing multi-scale feature map information through parallel work. Then, using the feature map information as input, the classification of each pixel is predicted through OCRNet network to achieve semantic segmentation of the image. The results show that: OCRNet-18 segmentation model is superior to other deep learning models in terms of PA, M IoU and Kappa coefficients, and the accuracy of PA (total), Kappa coefficient and M IoU are 0.913, 0.866, and 0.644 respectively.

Classification of soil is crucial for the agricultural domain as it is an essential task in geology and engineering domains. Various procedures are proposed to classify soil types in the literature, but many of them consumed much time or required specially designed equipments/applications. Classification of soil involves the accounting of various factors due to its diversified nature. It can be observed that several critical domain-oriented decisions often depend on the type of soil like farmers might be benefitted from knowing the kind of soil to choose crops accordingly for cultivation. We have employed different Convolution Neural Network (CNN) architectures to identify the soil type accurately in real-time. This paper describes the comparative evaluation in terms of performances of various CNN architectures, namely, ResNet50, VGG19, MobileNetV2, VGG16, NASNetMobile, and InceptionV3. These CNN models are used to classify four types of soils: Clay, Black, Alluvial, and Red. The performance of the ResNet50 model is the best with a training accuracy and training loss of 99.47% and 0.0252, respectively compared to other competing models considered in this paper.

Video surveillance is an important method to obtain the dynamic changes of green macroalgae along the coast. The paper proposes a coastal green macroalgae extraction method based on the SLIC superpixel segmentation, CNN and SVM to realize the automated recognition of green macroalgae from lots of high-resolution RGB video data collected by unmanned aerial vehicle (UAV) and handheld devices. Firstly, SLIC algorithm is used to generate the multi-scale patches on the original high-resolution image. Then, three classification CNN is used to divide the multi-scale patches into three types: green macroalgae, background and mixing. Finally, SVM algorithm is used to extract the green macroalgae to improve the accuracy at the pixel level in the mixed patches. In order to evaluate the performance of the proposed method, experiments are conducted on our coastal green macroalgae image dataset. Compared with the method of RGB vegetation indices (such as ExR, RGBVI, NGBDI), the overall accuracy (OA), F1 score, and Kappa of the green macroalgae extraction with the method proposed in this paper are up to 95.23%, 0.9612, 0.9436, respectively. The results show that our method is significantly better than that of RGB vegetation indices since it effectively reduces the influence of sea waves and light on the recognition results. The automated extraction method for coastal green macroalgae proposed in this paper can provide a reference for the automatic monitoring of coastal green macroalgae with high precision.

Pneumonia is one of the most common infectious diseases in clinic. X-ray chest is an important basis for early diagnosis of pneumonia. With the development of computer vision technology, using convolutional neural network to train pneumonia image classification model has been gradually applied to the process of medical clinical diagnosis. However, there are many problems in the process of using convolutional neural network to train pneumonia image classification model, such as too long model training time, over fitting and low accuracy due to too small training dataset. To solve these problems, this paper proposes an optimization method of pneumonia image classification model based on transfer learning and feature fusion, which is called Transfer Fusion. The Transfer Fusion optimization method will transplant the trained source model parameters to the target model, and add a specific feature fusion classification layer, so as to significantly shorten the training time of the new model, improve the accuracy and prevent over fitting. In this paper, Transfer Fusion optimization method is applied to three common convolutional neural network models: Google InceptionNetV3, MobileNetV2 and ResNet50. Through a large number of experiments, the performance of the three models has been significantly improved and improved.

The purpose of image enhancement is to improve the visual effect of the image. It is one of the hot spots of digital image enhancement technology to realize the enhancement effect by changing the contrast of gray scale. In this paper, a method of image contrast enhancement based on mean value segmentation is proposed, which is combined with pattern and image processing technology, to select the appropriate image gray membership function, and to select the threshold value of fuzzy theory and image processing by means of iteration and Otsu method, finally, the final image was decoded and the image was enhanced. In this paper, this technique is applied to the study of image enhancement of locomotive wheelset, and the image results of different processing methods are analyzed.

To solve the problems of texture blurring and structure inconsistency in large area image inpainting. This paper proposes an image inpainting method based on multi-scale feature fusion. We design a multi-scale feature fusion module to expand the receptive field. Besides, we devise attention module to capture information from distant areas in the feature map. In this paper, partial convolution and recursive structure are adopted to repair the boundary of the missing area progressively. This method can continuously strengthen the constraint on the center of the missing region and make the repair results more refined. Experimental results show that, compared with the existing image inpainting methods, this structure improves the performance of image inpainting, and the image quality generated by it is the best.

Artificial intelligence (AI) is the most looked up technology with a diverse range of applications across all the fields, whether it is intelligent transportation systems (ITS), medicine, healthcare, military operations, or others. One such application is autonomous vehicles (AVs), which comes under the category of AI in ITS. Vehicular Adhoc Networks (VANET) makes communication possible between AVs in the system. The performance of each vehicle depends upon the information exchanged between AVs. False or malicious information can perturb the whole system leading to severe consequences. Hence, the detection of malicious vehicles is of utmost importance. We use machine learning (ML) algorithms to predict the flaw in the data transmitted. Recent papers that used the stacking ML approach gave an accuracy of 98.44%. Decision tree-based random forest is used to solve the problem in this paper. We achieved accuracy and F1 score of 98.43% and 98.5% respectively on the VeRiMi dataset in this paper. Explainable AI (XAI) is the method and technique to make the complex black-box ML and deep learning (DL) models more interpretable and understandable. We use a particular model interface of the evaluation metrics to explain and measure the model’s performance. Applying XAI to these complex AI models can ensure a cautious use of AI for AVs.

This paper presents a hybrid technique of combining the BERT embedding method and the graph convolutional neural network. This combination is then employed to solve the text classification problem. Initially, we apply the BERT embedding method to the whole corpus in order to transform all the texts into numerical vectors. Then, the graph convolutional neural network will be applied to these numerical vectors to classify these texts into their appropriate classes. Especially, in our approach, we need only a few labeled texts for the model training. For the illustration, in this paper, we use the BBC news and the IMDB movie reviews datasets to perform our experiments, showing that the performance of the graph convolutional neural network model is better than the performances of the combination of the BERT embedding method with other classical machine learning models.

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 paper proposes a cognitive view on intonational contours aiming to describe their partitions in terms of elementary cognitive categories related to a generic information packaging (IPk) mechanism. We formulate the hypothesis that IPk structures pack auditory information items at the cortical level into relations that are marked at the utterance level by prosodic phrases. An IPk model based on this hypothesis is used in the paper for describing two pairs of contours that are presented in [1] as problematic for a categorical phonological description. The paper proposes a cognitive description of the respective contours after their partitioning into hierarchies of nested IPk units. In this view phonological events become marks of functional constituents at the cognitive level and the semantic differences between contours are reflected by their structural cognitive differences.

In three-dimensional reconstruction technology, it mainly includes the active time of flight method and the passive binocular stereo vision method. However, the time of flight method is easy to be affected by the external environment illumination and the corresponding device is expensive. The matching speed of the binocular stereo vision method is slow. Therefore, this paper chooses the line structured light method with good real-time performance, high accuracy and strong stability. The complex free-form surface is taken as the research object. Due to the serious mutual occlusion and high order derivability at any point, it is difficult to achieve complete 3D reconstruction on this surface. In this paper, a complete algorithm is designed for free-form surface reconstruction. Firstly, in order to solve the problem of beam discontinuity and sensitivity to strong light in the process of laser fringe center extraction, the gray gravity center method combining R-channel is proposed. Secondly, to solve the problems of the traditional optical plane calibration with few points and complex calibration process, a freely moving chessboard calibration method is proposed to improve the flexibility and accuracy of calibration. Then, the 3D point cloud is obtained and the surface is reconstructed using the measurement platform.

The robot needs to know the pose in order to maintain stability and execute the walking path. Current solutions either rely on visual data, which is easily affected by the environment and lighting conditions, or integrate kinematics and inertial measurement unit (IMU) measurement data, however, there will be drift problems caused by accumulated errors. Aiming at the defects and stability problems of vision sensor in location, this paper combines vision sensor and IMU to complete the high-precision pose estimation at low cost, designs the combined positioning algorithm based on the extended Kalman Filter (EKF). Specifically, this paper proposes the number of correctly matched feature points and depth error as the judgment conditions of the combined strategy, and uses the IMU data to construct a process model, merges the pose estimation results of the vision sensor, and selectively corrects the vision sensor. The robot positioning experiment was carried out in the indoor laboratory scene, results show that the algorithm proposed in this paper can effectively suppress the positioning stability problem of the vision sensor and improve the accuracy of the pose estimation.

This paper presents a method of wide angle fovea visual odometry (WAF-VO) for Wide Angle Fovea Sensor SLAM (WAF-SLAM), by which a unique locally-high accurate and wide-angle map is generated in addition to camera motion estimation. The WAF sensor is a special-made wide-angle sensor that is inspired from human visual function, i.e., the spatial resolution of the image is not uniform throughout the entire field of view (FOV); it is much higher in the central FOV and decreases rapidly towards the periphery. Our visual odometry method is strongly characterized by a wide-angle FOV and space-variant resolution of the input image from the WAF sensor. A locally-high accurate and wide-angle mapping method is proposed as a major part for WAF-SLAM together with the camera motion estimation. Our proposed method estimates camera motions more stably using very low-spatial-resolution wide-angle images remapped from the input image of the WAF sensor. Using the estimated camera motions, narrow-angle high accurate maps are generated from corresponding feature points in high-spatial resolution central regions of the input image. Wide-angle maps are generated from ones in middle-spatial-resolution wide-angle images remapped from the input image apart from the above very low-spatial-resolution images. When the wide-angle maps are generated, the number of extracted feature points is increased by adjusting contrast threshold values of SIFT feature according to regions of the FOV. A KNN matching method improved using epipolar constraint is proposed and employed for avoidance of mismatching the increased feature points. Thus, the wide-angle maps are generated from more correct corresponding feature points. Finally, the above two types of maps are combined into the unique locally-high accurate and wide-angle map, i.e., a WAF map. Using our proposed method, the WAF map was generated by verification experiments. Furthermore, the paper presents an evaluation of the accuracy and precision of the generated map.

Distance education has gained significance recently. However, its application to robotics education presents challenges as physical access to hardware is typically required. While educational robotics simulation platforms exist, most are limited in scope or do not facilitate remote collaboration well. This paper proposes a novel networked robotics simulator for education, where students are able to collaborate in a shared 3D virtual space their robots inhabit. The framework combines these simulated worlds with a block-based programming environment to enable distance robotics education for a wider audience. The users' programs run in a web browser on their computers and they issue commands to the virtual robots through a network protocol abstracted to simple-to-use blocks. Robots send back acknowledgements and sensor values through the same protocol. Interactive environments provide students with a more immersive educational experience, and allow for automatically evaluating student performance. In addition to sharing the virtual worlds remotely, students can also develop their programs together since the block-based environment supports both synchronous and asynchronous remote collaboration. The paper presents two scenarios showing a robotics challenge and an extension of the concept to a smart city scenario controlling traffic lights. Reduced barriers to entry for both robotics education and curriculum creation will allow for a more diverse set of students and course materials.