Search

Future wireless network technology will provide automobiles with a connectivity feature to consolidate the concept of vehicular networks that collaborate in conducting cooperative driving tasks. The full potential of connected vehicles, which promises road safety and a quality driving experience, can be leveraged if machine learning (ML) models guarantee robustness in performing core functions, including localization and controls. Location awareness, in particular, lends itself to the deployment of location-specific services and improvement of the operation performance. Localization entails direct communication to the network infrastructure, and the resulting centralized positioning solutions readily become intractable as the network scales up. As an alternative to the centralized solutions, this article addresses a decentralized principle of vehicular localization reinforced by ML techniques in dense urban environments with frequent inaccessibility to reliable measurement. As such, the collaboration of multiple vehicles enhances the positioning performance of ML approaches. A virtual testbed is developed to validate this ML model for real-map vehicular networks. Numerical results demonstrate the universal feasibility of cooperative localization, in particular, for dense urban area configurations. [ABSTRACT FROM AUTHOR]

In the field of wireless communication, transmitter localization technology is crucial for achieving accurate source tracking. However, the extant methodologies for localization face numerous challenges in wireless sensor networks (WSNs), particularly due to the constraints posed by the sparse distribution of sensors across large areas. We present DSLoc, a deep learning-based approach for transmitter localization in sparse WSNs. Our method is based on an improved high-resolution network model in neural networks. To address localization in sparse wireless sensor networks, we design efficient feature enhancement modules, and propose to locate transmitter locations in the heatmap using an image centroid-based method. Experiments conducted on WSNs with a 0.01% deployment density demonstrate that, compared to existing deep learning models, our method significantly reduces the transmitter miss rate and improves the localization accuracy by more than double. The results indicate that the proposed method offers more accurate and robust performance in sparse WSN environments. [ABSTRACT FROM AUTHOR]

Indoor localization is a critical task in many embedded applications, such as asset tracking, emergency response, and real-time navigation. In this letter, we propose a novel fingerprinting-based framework for indoor localization called SANGRIA that uses stacked autoencoder neural networks with gradient-boosted trees. Our approach is designed to overcome the device heterogeneity challenge that can create uncertainty in wireless signal measurements across embedded devices used for localization. We compare SANGRIA to several state-of-the-art frameworks and demonstrate 42.96% lower average localization error across diverse indoor locales and heterogeneous devices. [ABSTRACT FROM AUTHOR]

A real-time 3-D spatial localization system to track an unmanned aerial vehicle (UAV) using a single off-board conventional camera and an onboard circle-shaped colored light-emitting diode (LED) marker is presented. The image from the camera is color-segmented and then morphologically closed before being sent to the ellipse detector. The location of the drone is then estimated with geometric/optical calculations using the pixel coordinates provided by the real-time ellipse detector. The location data generated by this system was validated by a motion-capture system (MCS) that was simultaneously tracking the system in real time. The average position error was of 6 cm, while the processing speed achieved 15 locations/s, which are comparable with recent research references. The usage of a low-cost camera makes this method promising for most mobile UAV tracking applications. [ABSTRACT FROM AUTHOR]

Positron emission tomography is widely used in clinical and preclinical applications. Positronium lifetime carries information about the tissue microenvironment where positrons are emitted, but such information has not been captured because of two technical challenges. One challenge is the low sensitivity in detecting triple coincidence events. This problem has been mitigated by the recent developments of PET scanners with long (1-2 m) axial field of view. The other challenge is the low spatial resolution of the positronium lifetime images formed by existing methods that is determined by the time-of-flight (TOF) resolution (200-500 ps) of existing PET scanners. This paper solves the second challenge by developing a new image reconstruction method to generate high-resolution positronium lifetime images using existing TOF PET. Simulation studies demonstrate that the proposed method can reconstruct positronium lifetime images at much better spatial resolution than the limit set by the TOF resolution of the PET scanner. The proposed method opens up the possibility of performing positronium lifetime imaging using existing TOF PET scanners. The lifetime information can be used to understand the tissue microenvironment in vivo which could facilitate the study of disease mechanism and selection of proper treatments.

Unmanned Aerial Vehicles (UAVs) rely on satellite systems for stable positioning. However, due to limited satellite coverage or communication disruptions, UAVs may lose signals for positioning. In such situations, vision-based techniques can serve as an alternative, ensuring the self-positioning capability of UAVs. However, most of the existing datasets are developed for the geo-localization task of the objects captured by UAVs, rather than UAV self-positioning. Furthermore, the existing UAV datasets apply discrete sampling to synthetic data, such as Google Maps, neglecting the crucial aspects of dense sampling and the uncertainties commonly experienced in practical scenarios. To address these issues, this paper presents a new dataset, DenseUAV, that is the first publicly available dataset tailored for the UAV self-positioning task. DenseUAV adopts dense sampling on UAV images obtained in low-altitude urban areas. In total, over 27K UAV- and satellite-view images of 14 university campuses are collected and annotated. In terms of methodology, we first verify the superiority of Transformers over CNNs for the proposed task. Then we incorporate metric learning into representation learning to enhance the model’s discriminative capacity and to reduce the modality discrepancy. Besides, to facilitate joint learning from both the satellite and UAV views, we introduce a mutually supervised learning approach. Last, we enhance the Recall@K metric and introduce a new measurement, SDM@K, to evaluate both the retrieval and localization performance for the proposed task. As a result, the proposed baseline method achieves a remarkable Recall@1 score of 83.01% and an SDM@1 score of 86.50% on DenseUAV. The dataset and code have been made publicly available on https://github.com/Dmmm1997/DenseUAV. [ABSTRACT FROM AUTHOR]

This letter studies the impact of the ultra-wideband (UWB) sensor placement for a zero velocity update aided pedestrian inertial navigation solution that uses foot-to-foot ranging feedback to improve its localization accuracy. Several sensor configurations are investigated in both static and dynamic cases. The root mean square error is used as the performance metric for the static measurement accuracy, whereas the circular error probable is chosen to evaluate the sensor configurations in dynamic experiments.

In the field of wireless communication, transmitter localization technology is crucial for achieving accurate source tracking. However, the extant methodologies for localization face numerous challenges in wireless sensor networks (WSNs), particularly due to the constraints posed by the sparse distribution of sensors across large areas. We present DSLoc, a deep learning-based approach for transmitter localization in sparse WSNs. Our method is based on an improved high-resolution network model in neural networks. To address localization in sparse wireless sensor networks, we design efficient feature enhancement modules, and propose to locate transmitter locations in the heatmap using an image centroid-based method. Experiments conducted on WSNs with a 0.01% deployment density demonstrate that, compared to existing deep learning models, our method significantly reduces the transmitter miss rate and improves the localization accuracy by more than double. The results indicate that the proposed method offers more accurate and robust performance in sparse WSN environments.

The Polaris-LAMP multi-modal 3-D gamma-ray imager is a radiation mapping and imaging platform which uses a commercial off-the-shelf (COTS) detector integrated with a contextual sensor localization and mapping platform. The integration of these systems enables a free-moving radiation imaging capability with proximity mapping, coded-aperture, and Compton imaging modalities, which can create 3-D reconstruction of photon sources from tens of keV to several MeV. Gamma-ray events are recorded using a segmented cadmium zinc telluride (CZT) detector (Polaris-H Quad by H3D Inc., Ann Arbor, MI, USA), while scene data are derived from a contextual sensor and computation package developed by Lawrence Berkeley National Laboratory which includes GPS, laser ranging, and inertial measurement sensors. An onboard computer uses these inputs to create rapidly updating pose (10 Hz) and 3-D scene estimates using a simultaneous localization and mapping (SLAM) algorithm. The precise gamma-ray event location and timing resolution of the Polaris CZT sensor enables Compton imaging above several hundred keV, while photon sources at lower images are localized using coded-aperture imaging techniques. The multi-modal imaging concept enables imaging of diverse radiation sources spanning from the 59-keV emission of 241Am to the 1.1 and 1.3 MeV lines of 60Co. This work focuses on the description of the operational principles of the detector system and demonstrating the 3-D imaging performance in a variety of source detection and mapping scenarios. As a proof of concept, we demonstrate mapping complex environments, including both point source and distributed-source environments using proximity, coded-aperture, and Compton imaging modalities. Furthermore, we show the successful use of the system to perform measurements in high-background environments through analysis of arrays of uranium hexafluoride cylinders at the Paducah UF6 project site.

Prior works on text-based video moment localization focus on temporally grounding the textual query in an untrimmed video. These works assume that the relevant video is already known and attempt to localize the moment on that relevant video only. Different from such works, we relax this assumption and address the task of localizing moments in a corpus of videos for a given sentence query. This task poses a unique challenge as the system is required to perform: 2) retrieval of the relevant video where only a segment of the video corresponds with the queried sentence, 2) temporal localization of moment in the relevant video based on sentence query. Towards overcoming this challenge, we propose Hierarchical Moment Alignment Network (HMAN) which learns an effective joint embedding space for moments and sentences. In addition to learning subtle differences between intra-video moments, HMAN focuses on distinguishing inter-video global semantic concepts based on sentence queries. Qualitative and quantitative results on three benchmark text-based video moment retrieval datasets - Charades-STA, DiDeMo, and ActivityNet Captions - demonstrate that our method achieves promising performance on the proposed task of temporal localization of moments in a corpus of videos.

Automatic defect classification (ADC) systems automatically classify defects that inevitably occur during semiconductor manufacturing processes. ADC is the beginning of defect management that increases the yield of semiconductor chip production, and prevents accidents in the process. It takes a lot of engineer’s labor to classify defects, but ADC can be the answer to classify all defects at low cost. ADC employs the defect image of a wafer surface, captured using scanning electron microscopy (SEM). SEM images can feature a variety of backgrounds based on the defect position on the wafer and the process steps. The manual classification of SEM images involves significant labor costs related to hiring experienced engineers. Despite recent ADC studies reporting good performance, the lack of labeled images and various backgrounds make it difficult to apply ADC in actual manufacturing processes. To address this issue, automated defect classification with defect localization is proposed herein. To this end, a classification model is specifically designed for reducing the effect of varying backgrounds using defect localization. Defect localization uses an object detection model to provide the region information of defects in SEM images. We aimed to design a classification model and defect detection model using semi-supervised learning to reduce labeling costs. Experimental results indicate that the classification performance, over 15 classes, is improved by 12.56% (9.82%p), as compared with that of supervised models. [ABSTRACT FROM AUTHOR]

Automatic program repair has drawn more and more attention since software quality is facing increasing challenges. In existing approaches, the unlimited search space is considered to be the main limitation in finding the correct patch. So how to reduce the search space to improve the efficiency of automatic program repair remains a problem to be solved. In this work, we represent a similarity-based and location-awareness-based automatic program repair (SLARepair). SLARepair takes the similarity between codes as important search information. The search space is further subdivided by the location-awareness strategy to improve search efficiency. In addition, to better guide the search process, a new fitness function is designed for genetic programming, which brings notable improvements. Moreover, the patch verification time is further reduced by utilizing the test case prioritization approach combined with test case filtering. Extensive experiments demonstrate that our SLARepair outperforms the state-of-the-art approaches on the Defects4J benchmark and achieves competitive performances. [ABSTRACT FROM AUTHOR]

Emotion recognition based on electroencephalography (EEG) has attracted significant attention due to its wide range of applications, especially in Human-Computer Interaction(HCI). Previous research treats different segments of EEG signals uniformly, ignoring the fact that emotions are unstable and discrete during an extended period. In this paper, we propose a novel two-step spatial-temporal emotion recognition framework. First, considering that the human emotion has not only ”short-term continuity” but also ”long-term similarity”, we propose a hierarchical self-attention network to jointly model local and global temporal information, so as to localize most related segments and reduce the influence of noise at the temporal level. Second, in order to extract discriminative features at the spatial level to enhance the emotion recognition performance, we further employ the squeeze-and-excitation module (SE module) along with the channel correlation loss (CC-Loss) to select the most task-related channels. We also define a new task called emotion localization, which aims to localize fragments with stronger emotions. We evaluate the proposed method on the proposed emotion localization task and typical emotion recognition task with three publicly available datasets, i.e., SEED, DEAP, and MAHNOB-HCI. The experimental results demonstrate that the proposed approach outperforms state-of-the-art methods. [ABSTRACT FROM AUTHOR]

With the spread of tampered images, locating the tampered regions in digital images has drawn increasing attention. The existing tampering localization methods, however, suffer from severe performance degradation when the images are subjected to some post-processing, as the tampering traces would be distorted by the post-processing operations. The poor robustness against post-processing has become a bottleneck for the practical applications of image tampering localization techniques. In order to address this issue, this paper proposes a novel re storation-assisted framework for image tampering loc alization (ReLoc). The ReLoc framework mainly consists of an image restoration module and a tampering localization module. The key idea of ReLoc is to use the restoration module to recover a high-quality counterpart from the distorted tampered image, such that the distorted tampering traces can be re-enhanced, facilitating the tampering localization module to identify the tampered regions. To achieve this, the restoration module is optimized not only with the conventional constraints on image visual quality, but also with a forensics-oriented objective function. Furthermore, the restoration module and the localization module are trained alternately, which can stabilize the training process and is beneficial for improving the performance. The robustness of ReLoc has been evaluated by using several common post-processing operations, including lossy compressions, online social network transmission, and image resizing. Extensive experimental results show that ReLoc can significantly improve the localization performance compared to using a restoration-free model. In addition, we have shown that the restoration module in a well-trained ReLoc model is transferable for different localization modules and across different datasets. [ABSTRACT FROM AUTHOR]

In the underground parking lot (UPL) scenario, appearance changes in an indoor environment cause visual feature matching failures between features extracted from the image and features in the map, and further, lead to vehicle localization loss. In this article, a prior UPL mask is proposed for long-term vehicle localization. First, the UPL polygonal pattern is summarized based on structured environment characteristics. Then, feature matching is performed on the unannotated video data that reflect the degree of environmental change, and the polygonal pattern parameters are determined by a rule-based method using the statistical results of feature matching. Thus, the main distribution area of long-term static features in the mask is obtained quantitatively. Finally, the extraction of features in the mask for mapping and localization allows more static feature pairs to be matched for motion estimation. A real vehicle platform with stereo vision was set up for experimental verification. The comparison results demonstrated that the proposed method improved long-term vehicle localization stability and adaptability to overcome environmental changes. [ABSTRACT FROM AUTHOR]

Autonomous underwater vehicle (AUV) with towed magnetic anomaly detection (MAD) system is a reliable technique for detecting underwater magnetic targets. However, arranging too many magnetometers in the towed MAD system is not suitable, and the key lies in their reasonable arrangement. As a result, this letter proposes a new MAD system equipped with two scalar and one vector magnetometers, along with the corresponding multiparameter magnetic target localization method. By using two scalar magnetometers and the geometric relationship with vector magnetometer, the time when the vector magnetometer reaches the closest point approach (CPA) on the trajectory and the value range of the shortest distance are obtained. The value range of the shortest distance is cleverly set as the search boundary for the improved particle swarm optimization algorithm in this dimension, which greatly mitigates the randomness of the optimization algorithm and enhances the accuracy of the solution. Subsequently, the unique solution for the magnetic target parameters can be easily obtained by using the vector magnetometer. The combination of scalar and vector magnetometers proves to be both economical and effective. Numerical simulation and field experiments show that this method provides a new idea for static magnetic target detection. [ABSTRACT FROM AUTHOR]

Change detection (CD) plays an important role in Earth surface analysis. Current CD methods have achieved good performance in large flat areas, but CD of detailed parts is still a great challenge, and the loss of detail causes many faults around the change boundaries and on small objects. By analyzing the feature map of the widely used U-Net architecture in existing methods, we ascribe the detail loss to the depletion of detailed features during the top-to-down delivery in the U-Net architecture. The feature refine CD (FRCD) model is proposed in which the detection results are predicted directly from the multiscale features instead of the U-Net architecture. By direct prediction, the representation ability of details is enhanced, and thus the detection accuracy (Acc) of boundaries and small objects improves. Moreover, the normal upsampling in direct prediction is replaced with the deformable upsampling, which delivers detailed information from the low-level to the high-level via the deformable convolution, allowing the results to further fit boundaries in the FRCD model. Experimental results on two datasets confirm the effectiveness of FRCD compared to the state-of-the-art methods, and the CD results of boundaries and small objects are improved significantly by the proposed method. Code will be available after the acceptance of the letter in https://github.com/ijnokml/cdfr. [ABSTRACT FROM AUTHOR]

This letter introduces a tool for explaining the latent space generated by applying convolutional autoencoders (CAEs) to satellite image time series, entitled Grad-SLAM. We rely on backpropagated gradient interpretation combined with network activation localization. We use the proposed formula for multiple layers of the encoder, then scale and merge the results to generate a single date contribution metric for the generation of the latent space. We illustrate the potential of this method with the study of the unsupervised classification of agricultural Sentinel-1 time series. We show that critical characterizing dates for unsupervised retrieval of a given class are conditioned by the crop type’s radiometric signature and class count. We also present how Grad-SLAM can be used to enhance the understanding of unsupervised classification confusion. [ABSTRACT FROM AUTHOR]

Weakly supervised learning using image-level annotations has become a popular choice for reducing labeling efforts of remote sensing object extraction. Existing methods exploit inter-pixel relations within an individual image patch for object localizations. When facing large-scale remote sensing images, it is still challenging to obtain global semantic contexts across image patches for feature representation, resulting in inaccurate object localizations. To remedy these issues, we propose a local–global anchor guidance network (LGAGNet) for weakly supervised landslide extraction. Specifically, a structure-aware object locating (SOL) module is developed to capture the spatial structure of landslide objects and extract local category anchors containing informative feature embeddings. Furthermore, we leverage a global anchor aggregation (GAA) module to excavate semantic patterns across image patches based on a memory bank, which is then used as additional context cues to enhance the feature presentation through a cross-attention mechanism. Finally, a hybrid loss function is designed to guide the network training, considering category-aware semantic contrasts and local activation consistency. Experimental results on high-resolution aerial and satellite image datasets verify the effectiveness of the proposed approach on landslide extraction. [ABSTRACT FROM AUTHOR]

This letter pays attention to the mitigation of the radio frequency interference (RFI) sources located at coastal regions for synthetic aperture interferometric radiometer (SAIR). To remove an RFI at the visibility level, it will be necessary to estimate the brightness temperature (BT) of the point source, which is currently done by computing the median of the surrounded pixels to get the background BT and subtracting it from the original contaminated BT. However, this strategy is based on the assumption that the background BT distribution around the RFI is flat, and this is not reasonable for those RFIs from coastal cities and could result in estimation deviation. A robust RFI mitigation method is proposed without introducing complex computing. Instead of focusing only on the local area, the basic idea of our method is to step into the wide area and minimize the overall fluctuation of the pixels affected by the RFI in the BT image. The method tests based on SMOS visibility samples validate the correctness of the developed method. [ABSTRACT FROM AUTHOR]

Visual explanation of “black-box” models allows researchers in explainable artificial intelligence (XAI) to interpret the model’s decisions in a human-understandable manner. In this letter, we propose interpretable class activation mapping for tree crown detection (Crown-CAM) that overcomes inaccurate localization and computational complexity of previous methods while generating reliable visual explanations for the challenging and dynamic problem of tree crown detection in aerial images. It consists of an unsupervised selection of activation maps, computation of local score maps, and noncontextual background suppression to efficiently provide fine-grain localization of tree crowns in scenarios with dense forest trees or scenes without tree crowns. In addition, two intersection over union (IoU)-based metrics are introduced to effectively quantify both the accuracy and inaccuracy of generated explanations with respect to regions with or even without tree crowns in the image. Empirical evaluations demonstrate that the proposed Crown-CAM outperforms the Score-CAM, Augmented Score-CAM, and Eigen-CAM methods by an average IoU margin of 8.7, 5.3, and 21.7 (and 3.3, 9.8, and 16.5), respectively, in improving the accuracy (and decreasing inaccuracy) of visual explanations on the challenging NEON tree crown dataset. [ABSTRACT FROM AUTHOR]

The inadvertent incorporation of deviating samples into the measured indirect and direct path delays is generally unavoidable in the practical implementation of passive elliptic localization. These outlying observations, however, can do great harm to the positioning performance if left untreated. Here, a robust statistics-based method is put forward as the solution to such a problem. The non-outlier-resistant $\ell _{2}$ cost function in the traditional least squares (LS) formulation is replaced by a certain differentiable error measure that possesses resistance to the presence of abnormally large fitting errors. A globally optimized hybrid quasi-Newton and particle swarm optimization (PSO) algorithm is then developed for an efficient realization of the robust estimator. The strong capability of the presented approach to deal with outliers and its applicability to typical adverse localization environments are demonstrated via simulations. [ABSTRACT FROM AUTHOR]

Object detection in aerial images is a critical and essential task in the fields of geoscience and remote sensing. Despite the popularity of computer vision methods in detecting objects, these methods have been faced with significant limitations of aerial images such as appearance occlusion and variable object sizes. In this letter, we explore the limitations of conventional neck networks in object detection by analyzing information bottlenecks. We propose an enhanced neck network to address the information deficiency issue in current neck networks. Our proposed neck network, which serves as a bridge between the backbone network and the head network, comprises a global semantic network (GSNet) and a feature fusion refinement module (FRM). The GSNet is designed to perceive contextual surroundings and propagate discriminative knowledge through a bidirectional global pattern. The FRM is developed to exploit different levels of features to capture comprehensive location information. We validate the efficacy and efficiency of our approach through experiments conducted on two challenging datasets, DOTA and HRSC2016. Our method outperforms existing approaches in terms of accuracy and complexity, demonstrating the superiority of our proposed method. [ABSTRACT FROM AUTHOR]

This work considers the long baseline (LBL) acoustic localization problem. In the classical LBL localization approach, the time of arrival (TOA) of direct sound from each buoy is estimated independently, and the estimated TOAs are then utilized to solve for the target’s position. This approach ignores the fact that all of the estimated TOAs from different buoys correspond to the same target and thus are intrinsically correlated. To address this shortcoming, a novel localization framework based on track-before-detect (TBD) theory is established in this work, which allows the target location to be directly determined. Since it is completely freed from the traditional localization paradigm, some of the persistent and challenging problems faced by classical localization methods, such as direct sound selection and unacceptable false alarm/detection tradeoffs caused by threshold setting, can be circumvented. The effectiveness of the TBD approach is validated on both simulated and real datasets. The experimental results indicate that the proposed approach can track targets continuously and accurately and can be effectively implemented even under challenging conditions. [ABSTRACT FROM AUTHOR]

Underwater terrain image (UTI) matching provides a new idea for terrain-aided localization technology. However, the diversity of UTIs makes it very difficult to accurately match real-time terrain images (RTTIs) with reference terrain images (RTIs). To improve the localization accuracy of UTIs, this letter describes how UTI matching can be modeled as a deep learning problem based on supervised algorithms. Specifically, we propose a hierarchical matching strategy for the diverse features of UTIs by combining the deep and shallow feature characteristics of a convolutional neural network (CNN) to improve the localization accuracy. We also propose a sample-generation strategy based on candidate area quality index classification, and develop a data enhancement method by simulating the underwater terrain environment to overcome the lack of network training data. Simulations prove that the proposed method achieves accurate and stable localization results. [ABSTRACT FROM AUTHOR]

Reconfigurable intelligent surfaces (RISs) have tremendous potential for both communication and localization. While communication benefits are now well understood, the breakthrough nature of the technology may well lie in its capability to provide and support localization capabilities. We present an overview of RIS-enabled localization scenarios, considering various numbers of RISs, single- or multi-antenna base stations, narrowband or wideband transmissions, and near- and far-field operations. Based on this overview, we also highlight key research directions and open challenges specific to RIS-enabled localization and sensing. [ABSTRACT FROM AUTHOR]

Disasters are sudden and catastrophic events with fatal consequences. Time-sensitive information collection from disaster zones is crucial for improved and data-driven disaster response. However, information collection from disaster zones in a prompt way is not easy or even possible. Human-centric information provided by citizen sensors through social media platforms create an opportunity for prompt information collection from disaster zones. There is, nevertheless, limited scholarly work that provides a comprehensive review on the potential of social media analytics for disaster response. This study utilizes a systematic literature review with PRISMA protocol to investigate the potential of social media analytics for enhanced disaster response. The findings of the systematic review of the literature pieces (n = 102) disclosed that (a) social media analytics in the disaster management research domain is an emerging field of research and practice; (b) the central focus on the research domain is on the utilization of social media data for disaster response to natural hazards, but the social media data-driven disaster response to human-made disasters is an increasing research focus; (c) human-centric information intelligence provided by social media analytics in disaster response mainly concentrates on collective intelligence, location awareness, and situation awareness, and (d) there is limited scholarly research investigating near-real-time transport network management aftermath disasters. The findings inform authorities' decision-making processes as near-real time disaster response management depending on social media analytics is a critical element of securing sustainable cities and communities. [ABSTRACT FROM AUTHOR]

In this paper, we propose a three-dimensional autonomous drone exploration system (ADES) with a lightweight and low-latency saliency prediction model to explore unknown environments. Several studies have applied saliency prediction in drone exploration. However, these studies are not sufficiently mature. For example, the computational complexity and the size of the developed prediction models have not been considered. In addition, some studies have only proposed saliency prediction models without actually applying them to drones. The ADES system proposed in this paper has a small and fast saliency prediction model and uses a novel drone exploration approach based on visual-inertial odometry to solve the practical problems encountered during drone exploration, such as collisions with and the repeated exploration of salient objects. The proposed ADES system performs comparably to the state-of-the-art, multiple-discontinuous-image saliency prediction network TA-MSNet and enables drones to explore unknown environments with high efficiency. [ABSTRACT FROM AUTHOR]

Although facial landmark localization (FLL) approaches are becoming increasingly accurate in identifying facial components, one question remains unanswered: what is the impact of these approaches on subsequent, related tasks? In this paper, we focus on facial expression recognition (FER), where facial landmarks are used for face registration, which is a common usage. Since the common datasets for facial landmark localization do not allow for a proper measurement of performance according to the different difficulties (e.g., pose, expression, illumination, occlusion, motion blur), we also quantify the performance of recent approaches in the presence of head pose variations and facial expressions. Finally, we conduct a study of the impact of these approaches on FER. We show that the landmark accuracy achieved so far by optimizing the euclidean distance does not necessarily guarantee a gain in performance for FER. To deal with this issue, we propose a new evaluation metric for FLL that is more relevant to FER. [ABSTRACT FROM AUTHOR]

Fault localization is one of the most important issues for modular multilevel converters (MMCs) consisting of numerous switches. This article proposes an isolation forest (IF) based submodule (SM) switch open-circuit fault localization method for MMCs. Based on the continuous sampling SM capacitor voltages, a number of isolation trees (ITs) are produced to construct the IFs for MMCs. Through the comparison of continuous IFs’ outputs, the faulty SM can be effectively localized. The proposed IF-based fault localization method only requires SM capacitor voltages in the MMC to construct concise low-data-volume tree models, and uses sparsity and difference properties of outlier data to localize fault, and accordingly it simplifies calculation complexity. In addition, it does not require the MMC's mathematical models and manual setting of empirical thresholds. Simulation and experiment are conducted, and the results confirm the effectiveness of proposed method. [ABSTRACT FROM AUTHOR]

The malicious acoustic attack can easily cause the failure of micro-electro mechanical systems (MEMS) sensors. It can use ultrasonic waves to attack the MEMS gyroscope over long distances. This type of attack is highly hidden and harmful. Therefore, the purpose of this letter is to locate malicious sources and facilitate the eradication of the threat. We designed a prototype system to simulate an ultrasonic attack on an MEMS gyroscope, which can collect and analyze the state of the gyroscope after being attacked. Furthermore, we proposed $\sigma $ -standard deviation-based feature extraction, and then designed a neural-network-based traceability location method to seek for attacker’s location. The experimental results indicate that our approach can effectively and accurately locate the malicious acoustic source location. [ABSTRACT FROM AUTHOR]

The open-circuit fault detection and localization (FDL) technique can improve the reliability of the modular multilevel converter (MMC). However, the conventional software-based FDL methods usually have a heavy computation burden or a limited localization speed. This article proposes a simplified and fast software-based FDL approach for the grid-connected MMC. First, the errors between the measured state variables (the output current and the circulating current) and their estimated values are calculated. By comparing these errors with their threshold values, the switch fault can not only be detected but also be localized to the specific arm. Then, the capacitor voltages in this faulty arm are collected, and the submodule (SM) with the highest capacitor voltage is selected. To confirm the switch fault in this SM, a modified Pauta criterion is presented to check the abnormal voltage data. As a result, the computation burden of the proposed software-based FDL approach is significantly reduced, and the faulty SM can be localized in a short period. Simulation and experimental results verify that the proposed approach can effectively detect and localize different open-circuit faults, and it is immune to the step of power references. [ABSTRACT FROM AUTHOR]

Industrial Internet of Things (IIoTs) are significantly changing informative and manufacturing pattern in smart factories while it also brings security and trustworthiness issue. Concerning about trustworthiness issues and private preservation of tracking systems, a hierarchical framework with federated control theory is designed, which consists of a federated control center, network layer, and a federated control node. The framework combines a collaborative Cloud-Edge-End structure and machine learning-oriented localization, which further forms the EM-based federated scheme. On this basis, a trustworthy localization model is built with the untrustworthiness probability as a latent variable. By exploring expectation maximization (EM) of trustworthy localization, the local messages and aggression equations are derived in an iterative way of federated learning. The EM-based federated control scheme with machine learning-oriented localization is finally given. Experiments have been conducted to prove the localization accuracy and convergence of the proposed method with trustworthiness issue. The results show that trustworthy localization outperforms traditional methods without considering the security threats. [ABSTRACT FROM AUTHOR]

High-frequency activity (HFA) in intracranial electroencephalography recordings are diagnostic biomarkers for refractory epilepsy. Clinical utilities based on HFA have been extensively examined. HFA often exhibits different spatial patterns corresponding to specific states of neural activation, which will potentially improve epileptic tissue localization. However, research on quantitative measurement and separation of such patterns is still lacking. In this paper, spatial pattern clustering of HFA (SPC-HFA) is developed. The process is composed of three steps: (1) feature extraction: skewness which quantifies the intensity of HFA is extracted; (2) clustering: k-means clustering is applied to separate column vectors within the feature matrix into intrinsic spatial patterns; (3) localization: the determination of epileptic tissue is performed based on the cluster centroid with HFA expanding to the largest spatial extent. Experiments were conducted on a public iEEG dataset with 20 patients. Compared with existing localization methods, SPC-HFA demonstrates improvement (Cohen’s d $>0.2$) and ranks top in 10 out of 20 patients in terms of the area under the curve. In addition, after extending SPC-HFA to high-frequency oscillation detection algorithms, corresponding localization results also improve with effect size Cohen’s d $\geq 0.48$. Therefore, SPC-HFA can be utilized to guide clinical and surgical treatment of refractory epilepsy. [ABSTRACT FROM AUTHOR]

Anesthetic-induced loss of consciousness (LOC) has been studied using functional connectivity (FC) and functional network analysis (FNA), manifested as fragmentation of the whole-brain functional network. However, how the fragmented brain networks reversibly recover during the recovery of consciousness (ROC) remains vague. This study aims to investigate the changes in brain network structure during ROC, to better understand the network fragmentation during anesthesia, thus providing insights into consciousness monitoring. We analyzed EEG data recorded from 15 individuals anesthetized by sevoflurane. By investigating the properties of functional networks generated using different brain atlases and performing community detection for functional networks, we explored the changes in brain network structure to understand how fragmented brain networks recover during the ROC. We observed an overall larger FC magnitude during LOC than in the conscious state. The ROC was accompanied by the increasing binary network efficiency, decreasing FC magnitude, and decreasing community similarity with the functional atlas. Furthermore, we observed a negative correlation between modularity and community number ($\text{p} < 0.001$ and $\textit {BF}_{{10}} >4000$ , linear regression test), in which modularity increased and community number decreased during ROC. Our results show that a larger FC magnitude reveals excessive synchronization of neuronal activities during LOC. The increasing binary network efficiency, decreasing community number, and decreasing community similarity indicate the recovery of functional network integration. The increasing modularity implies the recovery of functional network segregation during ROC. The results suggest the limitation of FC magnitude and modularity in monitoring anesthetized states and the potential of integrated information theory to evaluate consciousness. [ABSTRACT FROM AUTHOR]

Remote sensing (RS) object detectors based on convolutional neural networks (CNNs) are hard to model global context dependencies. Transformer-based detectors can overcome this problem via global pairwise interactions, but more comprehensive information interactions are not systemically investigated to boost the detection performance. In view of this issue, we propose a transformer-guided multi-interaction network (TransMIN), which uses ResNet50-feature pyramid network (FPN) as the backbone for remote sensing object detection (RSOD). Specifically, we implement local–global feature interactions (LGFIs) by combining convolution and transformer in the residual blocks of ResNet50 to learn complementary features. We implement cross-view feature interactions (CVFIs) via transformers in the pyramid layers of FPN to capture the correlation between reference features (spatial edge priors and channel statistics) and pyramid features. This enhances edge information and suppresses background interference. In the detection head, we adopt a task-interactive sample assigner (TISA) by considering the interactions of classification and localization losses to obtain high-quality predictions. Experiments on two benchmark datasets demonstrate the superior detection performance of TransMIN over state-of-the-art methods. [ABSTRACT FROM AUTHOR]

Object detection in aerial images has achieved remarkable progress with the advent of deep convolutional neural networks (CNNs). It is, however, still a challenging task since the objects in aerial images are arbitrarily oriented and often densely packed. In this letter, we propose a novel method for oriented object detection in aerial images that represents objects as rotation equivariant semantic keypoints. Unlike conventional methods that represent object rotation according to angles from each axis in the Cartesian coordinate system, we represent object using a canonical orientation to ensure rotation equivariance. We accomplish this by representing an object as semantic keypoints, where each keypoint of the object consistently corresponds to the semantic part, regardless of rotation variation. To this end, we define the “head” point of the object as the canonical orientation and the remaining bounding box vectors as semantic keypoints in clockwise order. To discriminate visual attributes between different categories, we further use category-specific semantic keypoints, so that object classification and localization can be jointly solved in a cooperative manner. Our experiments demonstrate the effectiveness of rotation equivariant semantic keypoints on oriented object detection. [ABSTRACT FROM AUTHOR]

Audiovisual scenes are pervasive in our daily life. It is commonplace for humans to discriminatively localize different sounding objects but quite challenging for machines to achieve class-aware sounding objects localization without category annotations, i.e., localizing the sounding object and recognizing its category. To address this problem, we propose a two-stage step-by-step learning framework to localize and recognize sounding objects in complex audiovisual scenarios using only the correspondence between audio and vision. First, we propose to determine the sounding area via coarse-grained audiovisual correspondence in the single source cases. Then visual features in the sounding area are leveraged as candidate object representations to establish a category-representation object dictionary for expressive visual character extraction. We generate class-aware object localization maps in cocktail-party scenarios and use audiovisual correspondence to suppress silent areas by referring to this dictionary. Finally, we employ category-level audiovisual consistency as the supervision to achieve fine-grained audio and sounding object distribution alignment. Experiments on both realistic and synthesized videos show that our model is superior in localizing and recognizing objects as well as filtering out silent ones. We also transfer the learned audiovisual network into the unsupervised object detection task, obtaining reasonable performance. [ABSTRACT FROM AUTHOR]

Template-based discriminative trackers are currently the dominant tracking paradigm due to their robustness, but are restricted to bounding box tracking and a limited range of transformation models, which reduces their localization accuracy. We propose a discriminative single-shot segmentation tracker – D3S $_2$ 2 , which narrows the gap between visual object tracking and video object segmentation. A single-shot network applies two target models with complementary geometric properties, one invariant to a broad range of transformations, including non-rigid deformations, the other assuming a rigid object to simultaneously achieve robust online target segmentation. The overall tracking reliability is further increased by decoupling the object and feature scale estimation. Without per-dataset finetuning, and trained only for segmentation as the primary output, D3S $_2$ 2 outperforms all published trackers on the recent short-term tracking benchmark VOT2020 and performs very close to the state-of-the-art trackers on the GOT-10k, TrackingNet, OTB100 and LaSoT. D3S $_2$ 2 outperforms the leading segmentation tracker SiamMask on video object segmentation benchmarks and performs on par with top video object segmentation algorithms. [ABSTRACT FROM AUTHOR]

Weakly supervised temporal action localization aims at learning the instance-level action pattern from the video-level labels, where a significant challenge is action-context confusion. To overcome this challenge, one recent work builds an action-click supervision framework. It requires similar annotation costs but can steadily improve the localization performance when compared to the conventional weakly supervised methods. In this paper, by revealing that the performance bottleneck of the existing approaches mainly comes from the background errors, we find that a stronger action localizer can be trained with labels on the background video frames rather than those on the action frames. To this end, we convert the action-click supervision to the background-click supervision and develop a novel method, called BackTAL. Specifically, BackTAL implements two-fold modeling on the background video frames, i.e., the position modeling and the feature modeling. In position modeling, we not only conduct supervised learning on the annotated video frames but also design a score separation module to enlarge the score differences between the potential action frames and backgrounds. In feature modeling, we propose an affinity module to measure frame-specific similarities among neighboring frames and dynamically attend to informative neighbors when calculating temporal convolution. Extensive experiments on three benchmarks are conducted, which demonstrate the high performance of the established BackTAL and the rationality of the proposed background-click supervision. [ABSTRACT FROM AUTHOR]

Despite being widely used as a performance measure for visual detection tasks, Average Precision (AP) is limited in (i) reflecting localisation quality, (ii) interpretability and (iii) robustness to the design choices regarding its computation, and its applicability to outputs without confidence scores. Panoptic Quality (PQ), a measure proposed for evaluating panoptic segmentation (Kirillov et al., 2019), does not suffer from these limitations but is limited to panoptic segmentation. In this paper, we propose Localisation Recall Precision (LRP) Error as the average matching error of a visual detector computed based on both its localisation and classification qualities for a given confidence score threshold. LRP Error, initially proposed only for object detection by Oksuz et al. (2018), does not suffer from the aforementioned limitations and is applicable to all visual detection tasks. We also introduce Optimal LRP (oLRP) Error as the minimum LRP Error obtained over confidence scores to evaluate visual detectors and obtain optimal thresholds for deployment. We provide a detailed comparative analysis of LRP Error with AP and PQ, and use nearly 100 state-of-the-art visual detectors from seven visual detection tasks (i.e. object detection, keypoint detection, instance segmentation, panoptic segmentation, visual relationship detection, zero-shot detection and generalised zero-shot detection) using ten datasets to empirically show that LRP Error provides richer and more discriminative information than its counterparts. Code available at: https://github.com/kemaloksuz/LRP-Error. [ABSTRACT FROM AUTHOR]

Robust and real-time localization is an essential issue for autonomous driving. As a high-precision sensor, light detection and ranging (LiDAR) is widely used in autonomous driving. However, because LiDAR is vulnerable to environmental factors, such as fog, rain, and dynamic objects, robust localization based on LiDAR remains challenging. Moreover, the 3-D LiDAR is prohibitively expensive for consumer-grade applications. A robust and real-time outdoor localization method based solely on a single 2-D LiDAR is proposed to overcome these deficiencies. First, a grid matching algorithm based on data association is proposed to remove mismatches caused by great noises. Then, the Dempster–Shafer evidence theory-based strategy is proposed to fuse several frames of consecutive 2-D LiDAR data in a local window into a local map to eliminate dynamic objects. Finally, the generated local map is matched with the preconstructed global map to fulfill localization. Extensive experiments on the Karlsruhe Institute of Technology and Toyota Technological Institute at Chicago (KITTI) dataset validate the proposed method. It achieves high accuracy and real-time performance comparable and even superior to the 3-D LiDAR-based localization method. [ABSTRACT FROM AUTHOR]

With the advances in new technological trends and the reduction in prices of sensor nodes, wireless sensor networks (WSNs) and their applications are proliferating in several areas of our society, such as healthcare, industry, farming, and housing. Accordingly, in recent years, attention to localization has increased significantly since it is one of the main facets of a WSN. In a nutshell, localization is the process in which the position of a sensor node is retrieved by exploiting measurements from and between sensor nodes. Several techniques of localization have been proposed in the literature with different localization accuracies, complexities, and, hence, applicabilities. The localization accuracy is limited by fundamental limitations, theoretical and practical, which restrict the localization accuracy regardless of the technique employed in the localization process. In this article, we pay special attention to such limitations from theoretical and practical points of view and provide a comprehensive review of the state-of-the-art solutions that deal with such limitations. In addition, a discussion on the theoretical and practical limitations together with their recent solutions, remaining challenges, and perspectives is presented. [ABSTRACT FROM AUTHOR]

Time-of-flight-based localization requires high-accurate synchronization among the nodes in a network. Gradient clock synchronization (GCS) is a class of distributed algorithms capable of providing the demanded accuracy. Global clock rates defined by GCS algorithms are susceptible to drift relative to the individual hardware clock rates. This is due to the lack of a hard tie to a physical clock rate and is identified as the chaotic clock rate phenomenon. This scales range of measurements and can lead to stability issues in the network. This article presents a novel GCS algorithm for ultrawideband (UWB) ranging networks, addressing the chaotic global clock rate phenomenon. A correction term is introduced in the generic GCS algorithm so that the global clock rate is guaranteed to be converging into the average of individual clock rates. This also achieves asymptotic stability in the clock rate error state. The stability of the generic GCS and the proposed method for time-invariant hardware clock rates are compared using eigenvalue analysis in the clock error state space. A Kalman filter-based technique is used to precisely estimate the interanchor clock dynamics and ranges, which are then used in the asymptotic GCS (AGCS) update rule to calculate synchronization parameters. Simulations and experiments are conducted to evaluate the stability and synchronization accuracy of the proposed algorithm. The localization accuracy is evaluated for an indoor quadcopter localization task, which uses a range-assisted inertial navigation system (INS), resulting in rms position errors in the order of 0.2 m. [ABSTRACT FROM AUTHOR]

To simultaneously estimate the number of heads and locate heads with bounding boxes, we resort to detection-based crowd counting by leveraging RGB-D data and design a dual-path guided detection network (DPDNet). Specifically, to improve the performance of detection-based approaches for dense/tiny heads, we propose a density map guided detection module, which leverages density map to improve the head/non-head classification in detection network where the density implies the probability of a pixel being a head, and a depth-adaptive kernel that considers the variances in head sizes is also introduced to generate high-fidelity density map for more robust density map regression. In order to prevent dense heads from being filtered out during post-processing, we utilize such a density map for post-processing of head detection and propose a density map guided NMS strategy. Meanwhile, to improve the ability of detecting small heads, we also propose a depth-guided detection module to generate a dynamic dilated convolution to extract features of heads of different scales, and a depth-aware anchor is further designed for better initialization of anchor sizes in the detection framework. Then we use the bounding boxes whose sizes are generated with depth to train our DPDNet. Considering that existing RGB-D datasets are too small and not suitable for performance evaluation of data-driven based approaches, we collect two large-scale RGB-D crowd counting datasets, which comprise a synthetic dataset and a real-world dataset, respectively. Since the depth value at long-distance positions cannot be obtained in the real-world dataset, we further propose a depth completion method with meta learning, which fully utilizes the synthetic depth data to complete the depth value at long-distance positions. Extensive experiments on our proposed two RGB-D datasets and the MICC RGB-D counting dataset show that our method achieves the best performance for RGB-D crowd counting and localization. Further, our method can be easily extended to RGB image based crowd counting and achieves comparable or even better performance on the RGB datasets for both head counting and localization. [ABSTRACT FROM AUTHOR]