Search

Many sensors like digital fault indicators (DFIs) have been applied and promoted in distribution systems. The sensors can provide a technical mean for single‐line‐to‐ground (SLG) fault section location, but there are still some feature extraction and fault diagnosis problems. A novel SLG fault section location method utilizing auto‐encoder (AE) and fuzzy C‐means (FCM) clustering is presented in this work. Taking advantage of abundant information provided by DFIs, striking features can be extracted by the AE network, which is different from the artificially designed features that rely on prior knowledge. Compared with the learning‐based methods requiring massive training data, the proposed method only requires the data from one SLG fault. By applying the AE network to the zero‐sequence current measured by DFIs, the SLG fault section location's striking features could be obtained. Through feature classification by FCM clustering without setting threshold, the positional relationship between each detection node and the fault point would be distinguished to locate the fault section. Considering the abnormal communication of DFIs, the experiment proves that the proposed method can work effectively under various fault conditions. [ABSTRACT FROM AUTHOR]

Nowadays, smart monitoring devices such as digital fault indicator (DFI) have been installed in distribution systems to provide sufficient information for fault location. However, it is still a challenge to extract effective features from massive data for single-line-to-ground (SLG) fault-section location. This work proposes a novel method of fault-section location using a 1-D convolutional neural network (1-D CNN) and waveform concatenation. After SLG fault occurs, DFI measures the transient zero-sequence currents at double-ends of the line section, which could be concatenated to construct characteristic waveform. The features of characteristic waveforms would be extracted adaptively by 1-D CNN to locate the fault section. Furthermore, the problem where the on-site recorded data are hard to collect would be solved because 1-D CNN only needs a small number of samples for training in practical applications. The experimental results verified that the proposed method could work effectively under various fault conditions, even if a few DFIs are out of order. [ABSTRACT FROM AUTHOR]

Ocular surface damage is a major characteristic of dry eye syndrome. Ocular surface damage caused from dry eye refers to that there is superficial punctate keratitis (SPK), or also called the punctate dots, on the ocular surface. In the current diagnostic methods such as the ocular surface fluorescein-staining test, ophthalmologists dye the ocular surface to visualize punctate dots and then identify as well as count them for grading. Based on the grading results, ophthalmologists conduct a further diagnosis. For an expert, it is hard to achieve consistent results. Method: This study proposed to train a deep CNN (convolutional neural network) model to automatically detect punctate dots. Then we obtain a value, called the CNN-SPK value, which represents the coverage of punctate dots. Standard fluorescein-staining images from 101 participants were collected. Results: The correlation between the estimated CNN-SPK values of the rest 81 participants and the clinical grades were significant (r = 0.85; ${p} < 0.05$). Based on this observation, we suggest a statistical approach for the final grading. Using CNN-SPK values from 81 participants, as well as the corresponding clinical grades, we find CNN-SPK thresholds between any two adjacent grades. Also, we obtain the threshold between with- and without- SPK-symptoms, leading to 0.94 in sensitivity, and 0.79 in specificity. Conclusion: Our automatic method may be used to reliably grade the severity of punctate dots, to improve the efficiency of the dry diagnosis. [ABSTRACT FROM AUTHOR]

Due to the difficulty in locating high-resistance grounding faults, this paper proposes a novel fault location method for HVdc transmission lines by considering double-end unsynchronized using Hilbert–Huang transform and one-dimensional convolutional neural network (1D-CNN). After the fault signal is collected at both ends, the proposed method can achieve high-precision fault location, requiring only the two ends data transmission without time synchronization. After Empirical Mode Decomposition (EMD), the high-frequency components of the double-terminal fault signals are connected in series to make a characteristic waveform. This waveform contains characteristics of different fault types and distances, which can be learned by CNN. The trained CNN can then be used to achieve fault location effectively. As a comparison, two fault location methods based on traditional traveling wave and machine learning are introduced. Electromagnetic transient simulation software PSCAD/EMTDC has been used to carry out various types of fault simulation on the ± 500 kV HVdc transmission system. The results show that the proposed method can reliably and accurately locate line faults under fault resistance up to 5200 Ω. [ABSTRACT FROM AUTHOR]

Tear film instability is one of the major characteristics of dry eye syndrome. However, traditional diagnostic methods, such as the fluorescein tear film break-up time (FTBUT) test, are limited by the subjective interpretation of results. The test needs to manually identify break-up areas in the fluorescent image, thus producing variable diagnosis results. This paper proposes an automatic method to detect the fluorescent tear film break-up area using a deep convolutional neural network (CNN) model and to define its appearance as CNN-BUT. A digital slit-lamp recorded the standard FTBUT measurement for each of 80 study participants. Fifty participants were used to train the CNN model to identify the tear film break-up area, while the remaining 30 were used to validate the proposed CNN-BUT test. Among six normal controls and 24 dry eye patients enrolled in this paper, CNN-BUT was significantly lower in dry eye patients ( ${p} <0.05$ ). The correlation between CNN-BUT and FTBUT was also significant ( ${r} =0.9$ ; ${p} <0.05$ ). Using 5 s as the cutoff value, the CNN-BUT offered acceptable sensitivity and specificity to screen dry eye patients (0.83 and 0.95, respectively). These results indicate that CNN-BUT may be used to evaluate tear film stability and to assess the status of dry eye syndrome automatically. [ABSTRACT FROM AUTHOR]

Feature extraction for fault signals is critical and difficult in all kinds of fault detection schemes. A novel simple and effective method of faulty feeder detection in resonant grounding distribution systems based on the continuous wavelet transform (CWT) and convolutional neural network (CNN) is presented in this paper. The time-frequency gray scale images are acquired by applying the CWT to the collected transient zero-sequence current signals of the faulty feeder and sound feeders. The features of the gray scale image will be extracted adaptively by the CNN, which is trained by a large number of gray scale images under various kinds of fault conditions and factors. The features extraction and the faulty feeder detection can be implemented by the trained CNN simultaneously. As a comparison, two faulty feeder detection methods based on artificial feature extraction and traditional machine learning are introduced. A practical resonant grounding distribution system is simulated in power systems computer aided design/electromagnetic transients including DC, the effectiveness and performance of the proposed faulty feeder detection method is compared and verified under different fault circumstances. [ABSTRACT FROM PUBLISHER]

This work presents a low-noise amplifier (LNA) design with a wide-range gain control characteristic that integrates adjustable current distribution and output impedance techniques. For a given gain characteristic, the proposed LNA provides better wideband interference rejection performance than conventional LNA. Moreover, the proposed LNA also has a wider gain control range than conventional LNA. Therefore, it is suitable for satellite communications systems. The simulation results demonstrate that the voltage gain control range is between 14.5 and 34.2 dB for such applications (2600 MHz); the input reflection coefficient is less than −18.9 dB; the noise figure (NF) is 1.25 dB; and the third-order intercept point (IIP3) is 4.52 dBm. The proposed LNA consumes 23.85–28.17 mW at a supply voltage of 1.8 V. It is implemented by using TSMC 0.18-um RF CMOS process technology. [ABSTRACT FROM PUBLISHER]

Contact measurements of the respiratory rate using conventional electrocardiogram equipment requires patients to wear chest straps that can cause skin irritation and discomfort. Therefore, a real-time robust non-contact technique is developed for the measurement of respiratory rate variation. The changes of a simple harmonic motion between inhalation and exhalation from the human’s upper body can be observed from visual appearance. Therefore, in this paper, to characterize the motion, a salient region is automatically selected in the energy map resulted from Haar-like features through consecutive frames. Furthermore, a median optical flow signal is used to acquire the primary respiratory rate signal. The effective respiratory frequencies resulted from vertical motion variation are decomposed by median motion signal and then characterized by zero-crossing method before the frequencies were rectified by an estimation using the noise elimination methods. In the experiment, the performance is evaluated using an extensive data set obtained under distinct four respiratory types, regular respiration, respiration with body motion, respiration with distinct poses, and respiration with distinct capturing distances. Our approach develops a convenient non-contact method to evaluate the respiratory rate with the achievement of measurement under farther distance and also outperforms the current state-of-the-art approach in terms of high correlation coefficient. Therefore, the experiment results show its efficacy for real-world environment. [ABSTRACT FROM PUBLISHER]

Contact measurements of the cardiac pulse using the conventional electrocardiogram equipment requires patients to wear adhesive gel patches or chest straps that can cause skin irritation and discomfort. Commercially available pulse oximetry sensors that attach to the fingertips or earlobes also cause inconvenience for patients and the spring-loaded clips can be painful to use. Therefore, a novel robust non-contact technique is developed for the evaluation of heart rate variation. According to the periodic variation of reflectance strength resulting from changes to hemoglobin absorptivity across the visible light spectrum as heartbeats cause changes to blood volume in the blood vessels in the face, a reflectance signal is decomposed from consecutive frames of the green channel of the facial region. Furthermore, ensemble empirical mode decomposition of the Hilbert–Huang transform (HHT) is used to acquire the primary heart rate signal while reducing the effect of ambient light changes. The effective instantaneous frequencies from intrinsic mode functions decomposed by HHT are implemented by the multiple-linear regression model to evaluate heart rates before the frequencies were rectified by maximum likelihood method assuming Poisson distribution, and the minimum elapse time for heart rate evaluation is also evaluated in the estimate process. Experimental results show that our proposed approach provides a convenient non-contact method to evaluate heart rate and outperforms the current state-of-the-art method with higher accuracy and smaller variance. [ABSTRACT FROM PUBLISHER]

Rain removal from a single color image is a challenging problem as no temporal information among successive images can be obtained. In this paper, we propose a single-color-image-based rain removal framework by properly formulating rain removal as an image decomposition problem based on sparse representation. In our framework, an input color image is first decomposed into a low-frequency part and a high-frequency part by using the guided image filter so that the rain streaks would be in the high-frequency part with nonrain textures/edges, and the high-frequency part is then decomposed into a rain component and a nonrain component by performing dictionary learning and sparse coding. To separate rain streaks from the high-frequency part, a hybrid feature set, including histogram of oriented gradients, depth of field, and Eigen color, is employed to further decompose the high-frequency part. With the hybrid feature set applied, most rain streaks can be removed; simultaneously nonrain component can be enhanced. To the best of our knowledge, compared with the state-of-the-art approaches, the proposed method is among the first to focus on the problem of single color image rain removal and achieves promising results with not only the rain component being removed more completely, but also the visual quality of restored images being improved. [ABSTRACT FROM AUTHOR]

Speeded-Up Robust Features (SURF) is a robust and useful feature detector for various vision-based applications but it is unable to detect symmetrical objects. This paper proposes a new symmetrical SURF descriptor to enrich the power of SURF to detect all possible symmetrical matching pairs through a mirroring transformation. A vehicle make and model recognition (MMR) application is then adopted to prove the practicability and feasibility of the method. To detect vehicles from the road, the proposed symmetrical descriptor is first applied to determine the region of interest of each vehicle from the road without using any motion features. This scheme provides two advantages: there is no need for background subtraction and it is extremely efficient for real-time applications. Two MMR challenges, namely multiplicity and ambiguity problems, are then addressed. The multiplicity problem stems from one vehicle model often having different model shapes on the road. The ambiguity problem results from vehicles from different companies often sharing similar shapes. To address these two problems, a grid division scheme is proposed to separate a vehicle into several grids; different weak classifiers that are trained on these grids are then integrated to build a strong ensemble classifier. The histogram of gradient and SURF descriptors are adopted to train the weak classifiers through a support vector machine learning algorithm. Because of the rich representation power of the grid-based method and the high accuracy of vehicle detection, the ensemble classifier can accurately recognize each vehicle. [ABSTRACT FROM PUBLISHER]

State of the art video resizing methods usually produce perceivable visual discontinuities, especially in videos containing significant motion. To resolve the problem, contextual information about the focus of interest in consecutive video frames should be considered in order to preserve the visual continuity. In this paper, to detect the focus of interest with motion-tolerance, we propose a novel approach for modelling visual dynamics based on spatiotemporal slices (STS), which provide rich visual patterns along a large temporal scale. First, patch-based visual patterns are computed to generate a codebook of the automatically specified spatiotemporal extent determined by the contextual information in the STS. The codebook is then used to compute its associated response in each video frame, and eventually an importance map covering the focus of interest in a video clip can be obtained. To preserve the visual continuity of the content, particularly an important area, a multi-cue approach is used to guide a mesh-based non-homogeneous warping operation constrained by the trajectories in the STS. For the performance evaluation, we present a novel measure that utilizes patch-based Kullback–Leibler divergence (KL-divergence) to gauge the deformation of the focus of interest under the proposed video resizing approach. Experimental results show that the STS-based approach can generate retargeted videos effectively, while maintaining their isotropic manipulation and the continuous dynamics of visual perception. [ABSTRACT FROM PUBLISHER]

Modeling human crowds is an important issue for video surveillance and is a challenging task due to their unpredictable behavior. In this paper, the position of an isolated region that comprises an individual person or a set of occluded persons is detected by background subtraction. Each isolated region is considered a vertex and a human crowd is thus modeled by a graph. To construct a graph, Delaunay triangulation is used to systematically connect vertices and therefore the problem of event detection of human crowds is formulated as measuring the topology variation of consecutive graphs in temporal order. To effectively model the topology variations, local characteristics, such as triangle deformations and eigenvalue-based subgraph analysis, and global features, such as moments, are used and are finally combined as an indicator to detect if any anomalies of human crowd(s) present in the scene. Experimental results obtained by using extensive dataset show that our system is effective in detecting anomalous events for uncontrolled environment of surveillance videos. [ABSTRACT FROM PUBLISHER]

Given the rapid expansion of car ownership worldwide, vehicle safety is an increasingly critical issue in the automobile industry. The reduced cost of cameras and optical devices has made it economically feasible to deploy front-mounted intelligent systems for visual-based event detection for forward collision avoidance and mitigation. While driving at night, vehicles in front are generally visible by their taillights and brake lights. The brake lights are particularly important because they signal deceleration and potential collision. Therefore, in this paper, we propose a novel visual-based approach, based on the Nakagami-m distribution, for detecting brake lights at night by analyzing the taillights. Rather than using the knowledge of the heuristic features, such as the symmetry, position, and size of the rear-facing vehicle, we focus on finding the invariant features to model brake light scattering by Nakagami imaging and therefore conduct the detection process in a part-based manner. Experiments on an extensive data set show that our proposed system can effectively detect vehicle braking under different lighting and traffic conditions and thus demonstrates its feasibility in real-world environments. [ABSTRACT FROM PUBLISHER]

Advanced vehicle safety is a critical issue in recent years for automobiles, especially when the number of vehicles is growing rapidly worldwide. The decreasing cost of cameras makes it feasible to have an intelligent system of visually based event detection in front for forward collision avoidance and mitigation. When driving at night, vehicles in front are generally visible by their taillights. The brake lights are particularly important because drivers need to focus on them. Therefore, in this paper, we propose a novel approach that can detect brake lights at night using a camera by analyzing the signal in both spatial and frequency domains. Unlike the traditional approaches that employ the knowledge of the heuristic features, such as symmetry and position of rear facing vehicle, size, and so on, we focus on finding the invariant features from the regions of brake lights in the frequency domain and therefore can conduct the detection process in a part-based manner. Experiments with an extensive dataset show that our proposed system can efficiently and effectively detect brake lights under different lighting and traffic conditions, and thus prove its feasibility in real-world environments. [ABSTRACT FROM PUBLISHER]

This paper introduces a visual-based system, which can count the number of viewers and recognize their gender in front of an electronic billboard in real-time video streams. The viewers actually watching an advertisement are captured via frontal face detection techniques. To count the number of viewer precisely, the problem of occlusions between viewers is tackled. Besides, a complementary set of features is extracted from the torso of a viewer due to the fact that the part of the body contains relatively rich discriminative information than other body parts. In addition, for conducting robust viewer recognition, an online classifier trained by AdaBoost is developed. To recognize the gender of the counted viewers, an approach based on spatiotemporal probabilistic framework is proposed. Our experimental results demonstrate the robustness of the proposed system for the viewer counting and gender recognition tasks. [ABSTRACT FROM AUTHOR]

Gender recognition is a challenging task in real life images and surveillance videos due to their relatively lowresolution, under uncontrolled environment and variant viewing angles of human subject. Therefore, in this paper, a system of real-time gender recognition for real life images is proposed. The contribution of this work is three-fold. In order to make the system robust, a mechanism of decision making based on the combination of surrounding face detection, context-regions enhancement and confidence-based weighting assignment is designed. Experiment results obtained by using extensive dataset show that our system is effective and efficient in recognizing genders for uncontrolled environment of real life images. [ABSTRACT FROM PUBLISHER]