Deep neural networks (DNN) have achieved unprecedented success in numerous machine learning tasks in various domains. However, the existence of adversarial examples raises our concerns in adopting deep learning to safety-critical applications. As a result, we have witnessed increasing interests in studying attack and defense mechanisms for DNN models on different data types, such as images, graphs and text. Thus, it is necessary to provide a systematic and comprehensive overview of the main threats of attacks and the success of corresponding countermeasures. In this survey, we review the state of the art algorithms for generating adversarial examples and the countermeasures against adversarial examples, for three most popular data types, including images, graphs and text.
In the last two decades, improvements in materials, sensors and machine learning technologies have led to a rapid extension of electronic nose (EN) related research topics with diverse applications. The food and beverage industry, agriculture and forestry, medicine and health-care, indoor and outdoor monitoring, military and civilian security systems are the leading fields which take great advantage from the rapidity, stability, portability and compactness of ENs. Although the EN technology provides numerous benefits, further enhancements in both hardware and software components are necessary for utilizing ENs in practice. This paper provides an extensive survey of the EN technology and its wide range of application fields, through a comprehensive analysis of algorithms proposed in the literature, while exploiting related domains with possible future suggestions for this research topic.
Social media is a platform to express one′s views and opinions freely and has made communication easier than it was before. This also opens up an opportunity for people to spread fake news intentionally. The ease of access to a variety of news sources on the web also brings the problem of people being exposed to fake news and possibly believing such news. This makes it important for us to detect and flag such content on social media. With the current rate of news generated on social media, it is difficult to differentiate between genuine news and hoaxes without knowing the source of the news. This paper discusses approaches to detection of fake news using only the features of the text of the news, without using any other related metadata. We observe that a combination of stylometric features and text-based word vector representations through ensemble methods can predict fake news with an accuracy of up to 95.49%.
Hyperspectral images in remote sensing include hundreds of spectral bands that provide valuable information for accurately identify objects. In this paper, a new method of classifying hyperspectral images using spectral spatial information has been presented. Here, using the hyperspectral signal subspace identification (HYSIME) method which estimates the signal and noise correlation matrix and selects a subset of eigenvalues for the best representation of the signal subspace in order to minimize the mean square error, subsets from the main sample space have been extracted. After subspace extraction with the help of the HYSIME method, the edge-preserving filtering (EPF), and classification of the hyperspectral subspace using a support vector machine (SVM), results were then merged into the decision-making level using majority rule to create the spectral-spatial classifier. The simulation results showed that the spectral-spatial classifier presented leads to significant improvement in the accuracy and validity of the classification of Indiana, Pavia and Salinas hyperspectral images, such that it can classify these images with 98.79%, 98.88% and 97.31% accuracy, respectively.
This paper proposes a solution to localization and classification of rice grains in an image. All existing related works rely on conventional based machine learning approaches. However, those techniques do not do well for the problem designed in this paper, due to the high similarities between different types of rice grains. The deep learning based solution is developed in the proposed solution. It contains pre-processing steps of data annotation using the watershed algorithm, auto-alignment using the major axis orientation, and image enhancement using the contrast-limited adaptive histogram equalization (CLAHE) technique. Then, the mask region-based convolutional neural networks (R-CNN) is trained to localize and classify rice grains in an input image. The performance is enhanced by using the transfer learning and the dropout regularization for overfitting prevention. The proposed method is validated using many scenarios of experiments, reported in the forms of mean average precision (mAP) and a confusion matrix. It achieves above 80% mAP for main scenarios in the experiments. It is also shown to perform outstanding, when compared to human experts.
Orientation measurement of objects is vital in micro assembly. In this paper, we present a novel method based on monocular microscopic vision for 3-D orientation measurement of objects with planar surfaces. The proposed methods aim to measure the orientation of the object, which does not require calibrating the intrinsic parameters of microscopic camera. In our methods, the orientation of the object is firstly measured with analytical computation based on feature points. The results of the analytical computation are coarse because the information about feature points is not fully used. In order to improve the precision, the orientation measurement is converted into an optimization process base on the relationship between deviations in image space and in Cartesian space under microscopic vision. The results of the analytical computation are used as the initial values of the optimization process. The optimized variables are the three rotational angles of the object and the pixel equivalent coefficient. The objective of the optimization process is to minimize the coordinates differences of the feature points on the object. The precision of the orientation measurement is boosted effectively. Experimental and comparative results validate the effectiveness of the proposed methods.
Tracking registration is a key issue in augmented reality applications, particularly where there are no artificial identifier placed manually. In this paper, an efficient markerless tracking registration algorithm which combines the detector and the tracker is presented for the augmented reality system. We capture the target images in real scenes as template images, use the random ferns classifier for target detection and solve the problem of reinitialization after tracking registration failures due to changes in ambient lighting or occlusion of targets. Once the target has been successfully detected, the pyramid Lucas-Kanade (LK) optical flow tracker is used to track the detected target in real time to solve the problem of slow speed. The least median of squares (LMedS) method is used to adaptively calculate the homography matrix, and then the three-dimensional pose is estimated and the virtual object is rendered and registered. Experimental results demonstrate that the algorithm is more accurate, faster and more robust.
This paper presents a two-stage smooth-optimal trajectory tracking strategy. Different from existing methods, the optimal trajectory tracked point can be directly determined in an uncalibrated fish-eye image. In the first stage, an adaptive trajectory tracking controller is employed to drive the tracking error and the estimated error to an arbitrarily small neighborhood of zero. Afterwards, an online smooth-optimal trajectory tracking planner is proposed, which determines the tracked point that can be used to realize smooth motion control of the mobile robot. The tracked point in the uncalibrated image can be determined by minimizing a utility function that consists of both the velocity change and the sum of cross-track errors. The performance of our planner is compared with other tracked point determining methods in experiments by tracking a circular trajectory and an irregular trajectory. Experimental results show that our method has a good performance in both tracking accuracy and motion smoothness.
The aim of this work is to model and analyze the behavior of a new smart nano force sensor. To do so, the carbon nanotube has been used as a suspended gate of a metal-oxide-semiconductor field-effect transistor (MOSFET). The variation of the applied force on the carbon nanotube (CNT) generates a variation of the capacity of the transistor oxide-gate and therefore the variation of the threshold voltage, which allows the MOSFET to become a capacitive nano force sensor. The sensitivity of the nano force sensor can reach 0.124 31 V/nN. This sensitivity is greater than results in the literature. We have found through this study that the response of the sensor depends strongly on the geometric and physical parameters of the CNT. From the results obtained in this study, it can be seen that the increase in the applied force increases the value of the MOSFET threshold voltage VTh. In this paper, we first used artificial neural networks to faithfully reproduce the response of the nano force sensor model. This neural model is called direct model. Then, secondly, we designed an inverse model called an intelligent sensor which allows linearization of the response of our developed force sensor.
Based on compressive sensing and fractional discrete cosine transform (DCT) via polynomial interpolation (PI-FrDCT), an image encryption algorithm is proposed, in which the compression and encryption of an image are accomplished simultaneously. It can keep information secret more effectively with low data transmission. Three-dimensional piecewise and nonlinear chaotic maps are employed to obtain a generating sequence and the exclusive OR (XOR) matrix, which greatly enlarge the key space of the encryption system. Unlike many other fractional transforms, the output of PI-FrDCT is real, which facilitates the storage, transmission and display of the encrypted image. Due to the introduction of a plain-image-dependent disturbance factor, the initial values and system parameters of the encryption scheme are determined by cipher keys and plain-image. Thus, the proposed encryption scheme is very sensitive to the plain-image, which makes the encryption system more secure. Experimental results demonstrate the validity and the reliability of the proposed encryption algorithm.
To ensure the safe operation of any software controlled critical systems, quality factors like reliability and safety are given utmost importance. In this paper, we have chosen to analyze the impact of logic error that is one of the contributors to the above factors. In view of this, we propose a novel framework based on a data driven approach known as software failure estimation with logic error (SFELE). Here, the probabilistic nature of software error is explored by observing the operation of a safety critical system by injecting logic fault. The occurrence of error, its propagations and transformations are analyzed from its inception to end of its execution cycle through the hidden Markov model (HMM) technique. We found that the proposed framework SFELE supports in labeling and quantifying the behavioral properties of selected errors in a safety critical system while traversing across its system components in addition to reliability estimation of the system. Our attempt at the design level can help the design engineers to improve their system quality in a cost-effective manner.