Tracking Network Research Articles

Teacher-student learning of generative adversarial network-guided diffractive neural networks for visual tracking and imaging

Teacher-student learning of generative adversarial network-guided diffractive neural networks for visual tracking and imaging

Latent factors underlying the symptoms of adult-onset myotonic dystrophy type 1 during the clinical course

BackgroundMyotonic dystrophy type 1 (DM1) is a multisystem genetic disorder that classically presents with symptoms associated with myotonia, early onset cataracts, and muscular weakness, although the presentation and pattern of disease progression is quite varied. Presenting symptoms are well documented among adults with DM1. However, less is known about the co-occurrence of symptoms over time. We aimed to use factor analysis to explore the correlation pattern of signs and symptoms (S/S) that emerged during the clinical course.ResultsClinical records of 228 individuals with adult onset DM1 were abstracted using the Muscular Dystrophy Surveillance, Tracking, and Research Network (MD STARnet) from a six-site cohort in the United States during an eight-year study period. Factor analysis was used to group the correlated S/S into latent factors. Three factors were identified. Group 1: ‘Facial Weakness/Myotonia’ includes the two most common S/S, as indicated by its name. Group 2: ‘Skeletal Muscle Weakness’ includes eight muscular S/S and is more frequently reported by males and those with older age at onset. Group 3: ‘Gastrointestinal distress/Sleepiness’ includes four non-muscular S/S and hand stiffness. The abstracted medical records reported that over 63% of individuals had S/S from all three groups. Associations of covariates with factor scores were also examined using linear regression. CTG repeat length was significantly positively associated with higher factor scores for all three factors.ConclusionsThis study identified three latent factors of S/S which accumulated during the clinical course of adult onset DM1.

Respiratory function and evaluation in individuals with facioscapulohumeral muscular dystrophy in the Muscular Dystrophy Surveillance, Tracking and Research Network

Respiratory function and evaluation in individuals with facioscapulohumeral muscular dystrophy in the Muscular Dystrophy Surveillance, Tracking and Research Network

The INPHET- a global approach to Environmental Public Health Tracking

Abstract Environment and health monitoring and information systems are fundamental infrastructure for evidence provision for policy making. The International Network for Public Health and Environment Tracking (INPHET) was formed by a community of environmental public health practitioners to address common interests and activities that advance environmental public health surveillance (EPHS) and data systems, and their uses to inform public health actions. Its main goal is to address the vast inequalities across countries and communities in their ability to monitor, assess, inform and mitigate environmental and health issues. INPHET supports the relevant role of environmental epidemiology as the pillar science providing evidence for decision making and interventions aiming at reducing and preventing environmental related diseases and contributes to an overall vision for public health. As a “Community of Practice”, INPHET focuses on sharing experiences and expertise; addressing common challenges for implementing and enhancing the science of EPHS/tracking; improving understanding of the link between the environment and health; and providing the appropriate evidence that informs public policy. The network, initiated in conference meetings in 2012-13 and formally launched in Modena, Italy in 2014, has carried out several activities which aim at sharing expertise and learning from different models of environmental and health data integration and surveillance systems that directly inform policy. These include Twinning cooperation with Euro-Asian and Balkan countries and sharing of expertise with African Public Health Agencies. Some experiences are presented in this workshop and highlight the EPHS/EPHT model adopted by several national public health agencies. It promotes that such integration and uses of environmental and health data is key for EPHS, including their accessibility and usability for communities of public health researchers, practitioners and policy makers.

EHT and environmental justice at EPA Victoria: Aspirations and regulatory realities

Abstract EPA Victoria (Australia) is a statutory authority that regulates polluting activities and advises other Victorian government departments on relevant policy settings. In 2016, a public health function was installed at EPA Vic to provide emergency health messaging during pollution events, review research evidence of environmental-health relationships and assess industry-contracted health risk assessments of polluting entities. An Environmental Health Tracking Network (EHTN) is under development to support these activities. The EHTN is an agile data and analytical infrastructure for translating expertise across an expanding multidisciplinary network of public health researchers, EPA operational staff and public sector policy makers. Its current interactive dashboards calculate health impacts and associated costs of air pollution scenarios, model land-use influences on land and water contamination and identify coincidences of sociodemographic-, urban environment- and pollution exposure-disadvantage in a high spatial resolution environmental justice (EJ) framework. Regulatory leverage of EHTN intelligence, however, requires legislated EJ thresholds and/or legally tractable costing of industry-externalised human health impacts. Thus, to establish triggers for more stringent conditions on polluting activities and promote equity of interventions and regulations, novel regionally and EJ specific health risk functions are required. Considering these ambitions and challenges, the EHTN at EPA Vic is focussed on 1) identifying and characterising vulnerable populations at high spatial resolution, 2) estimating associated exposure-health effect size modifications using primary epidemiological methods and 3) applying these to health impact and cost assessments of pollution and waste. We seek defensible grounds on which to impose more stringent standards for polluters and redress entrenched social, health and economic disadvantage. Speakers/Panelists Hanna Tolonen THL, Helsinki, Finland

CLOWN: The PASO Cloud Detection for Optimization of Automatic Optical Surveys

Orbiting space objects have become in the last decade a major nuisance impacting ground astronomy and orbiting space assets, from observatories to satellites and space stations. In particular with the rise of the satellite population in Low Earth Orbits, space objects are becoming an even bigger threat and a strong problem to astronomical observations. To tackle these threats, several coordinated surveillance networks composed of dedicated sensors (telescopes, radars, and laser ranging facilities) track and survey space objects, from debris to active satellites. As part of the European Space Surveillance & Tracking network, Portugal is developing the Pampilhosa da Serra Space Observatory, with both radio and optical telescopes dedicated to the Space Situational Awareness domain, deployed at a Dark Sky destination. To optimize telescope survey time, we developed CLOud Watcher at Night (CLOWN), an application interface that automatically monitors clouds in real time. This software can correctly trace cloud positions in the sky and provide accurate pointing information to the observation planning of the optical telescope to avoid cloudy areas. CLOWN only requires the use of an all-sky camera, which is already a norm in observatories with optical telescopes and can be used with any camera, including those for which no information about its model specification do exist. CLOWN does not require great computing power, and it does not require the installation of additional equipment. CLOWN results are very promising and confirm that the app can correctly identify clouds in a variety of different conditions and cloud types.

Design of a robust intelligent controller based neural network for trajectory tracking of high-speed wheeled robots

The control design of wheeled mobile robots is often accomplished based on the robot’s kinematics which imposes critical challenges in the motion tracking control of such systems. In the related literature, most works designed dynamic controllers based on the terms of voltage or torque; however, the velocity is often utilized in industrial and commercial applications. To address this issue, this paper focuses on the design of a velocity-based control for the trajectory-tracking problem of mobile robots in practical applications including measurement acquisition. The control design of the mobile robot is realized in two separate parts using kinematic control and dynamic control. The design of kinematic control is carried out based on the robot’s kinematics where the motion is described without considering the forces and torques. A radial basis function (RBF) neural network (NN) based on the model-free controller is designed for the dynamic controller of mobile robots without relying on the system dynamics. In the proposed dynamic control, a time-delay estimation is adopted to estimate the disturbances and noises in the mobile robot and remove them from the feedback loop control. Several typical scenarios of mobile robots with high-speed movements are conducted to assess the feasibility of the proposed NN controller-based time-delay estimation under noises and disturbances. To show the supremacy of the suggested controller, the dynamic responses of the mobile robot test system are compared with the prevalent controllers. The extensive simulation and real-time examinations reveal the capability of the proposed NN controller-based time-delay estimation to control high-speed mobile robots under high-level noises and disturbances.

Self-Attention Factor Graph Neural Network for Multiagent Collaborative Target Tracking

Self-Attention Factor Graph Neural Network for Multiagent Collaborative Target Tracking

Multi-attention associate prediction network for visual tracking

Classification-regression prediction networks have realized impressive success in several modern deep trackers. However, there is an inherent difference between classification and regression tasks, so they have diverse even opposite demands for feature matching. Existed models always ignore the key issue and only employ a unified matching block in two task branches, decaying the decision quality. Besides, these models also struggle with decision misalignment situation. In this paper, we propose a multi-attention associate prediction network (MAPNet) to tackle the above problems. Concretely, two novel matchers, i.e., category-aware matcher and spatial-aware matcher, are first designed for feature comparison by integrating self, cross, channel or spatial attentions organically. They are capable of fully capturing the category-related semantics for classification and the local spatial contexts for regression, respectively. Then, we present a dual alignment module to enhance the correspondences between two branches, which is useful to find the optimal tracking solution. Finally, we describe a Siamese tracker built upon the proposed prediction network, which achieves the leading performance on five tracking benchmarks, consisting of LaSOT, TrackingNet, GOT-10k, TNL2k and UAV123, and surpasses other state-of-the-art approaches.

AMTrack:Transformer tracking via action information and mix-frequency features

Nowadays, Transformer-based visual tracking algorithms have been developing quickly because of the self-attention mechanism of Transformer, which has the capability to model global information. Although the self-attention mechanism in Transformer can effectively capture long-range dependencies in feature space, they only use flattened two-dimensional features and are unable to capture long-range temporal dependencies. Furthermore, since the self-attention in Transformer functions as a low-pass filter, it picks up on low-frequency features of the target while ignoring high-frequency features. This research suggests a Transformer tracker based on action information and mix-frequency features (AMTrack) to address these problems. Specifically, to address the lack of temporal remote dependencies, we introduce the target action aware module and the target action offset module. The target action aware module sets up several pathways to extract spatio-temporal, channel, and motion feature independently. In contrast, the target action offset module computes the target’s offset information by computing relative feature maps. Furthermore, in order to address the imbalance between high and low frequency features, we propose a mix-frequency attention and multi-frequency self-attention convolutional block. The mix-frequency attention uses high-frequency features within partitioned local windows as input for the high-frequency branch and average-pooled low-frequency features as the input for the low-frequency branch, calculating attention scores in both branches respectively. The multi-frequency self-attention convolutional block uses self-attention to capture low-frequency features and convolution to capture high-frequency features. Extensive experiments are carried out on eight challenging tracking datasets (e.g., OTB100 (Object Tracking Benchmark 100), NFS (Need For Speed), UAV123 (Unmanned Aerial Vehicles 123), TC128 (Temple Color 128), VOT2018 (Visual Object Tracking 2018), LaSOT (Large-scale Single Object Tracking), TrackingNet (Tracking Network), GOT-10k (Generic Object Tracking-10k)), and the experimental results show that our tracker achieves excellent tracking performance when compared with several state-of-the-art tracking algorithms. The experimental results show that on LaSOT, the success rates AUC (Area Under Curve), PNorm, and P reach 65.8%, 69.2%, and 68.0%, respectively, where the AUC value is 2.1% higher than the baseline algorithm TrDiMP (Transformer Discriminative Model Prediction). On other datasets, our tracker also achieves excellent tracking performance.

SiamATA: an asymmetric target-aware and frequency domain task-aware Siamese network for visual tracking

SiamATA: an asymmetric target-aware and frequency domain task-aware Siamese network for visual tracking

Hyperspectral Attention Network for Object Tracking

Hyperspectral video provides rich spatial and spectral information, which is crucial for object tracking in complex scenarios. Despite extensive research, existing methods often face an inherent trade-off between rich spectral information and redundant noisy information. This dilemma arises from the efficient utilization of hyperspectral image data channels. To alleviate this problem, this paper introduces a hierarchical spectral attention network for hyperspectral object tracking. We employ a spectral band attention mechanism with adaptive soft threshold to examine the correlations across spectral bands, which integrates the information available in various spectral bands and eliminates redundant information. Moreover, we integrate spectral attention into a hierarchical tracking network to improve the integration of spectral and spatial information. The experimental results on entire public hyperspectral competition dataset WHISPER2020 show the superior performance of our proposed method compared with that of several related methods in visual effects and objective evaluation.

Development of autonomous navigation system based on neural network and visual servoing for row-crop tracking in vegetable greenhouses

Accurate navigation line localization strategy and robust path tracking control scheme are two key techniques for advancing precision agriculture and enhancing the sustainability of smart farms. In this paper, a superior and cost-effective autonomous navigation system for a visual servoing-based agricultural vehicle in the greenhouse scenario with a distinct layout of two crop rows in each ridge is presented. First, based on image segmentation techniques, a neural network model named DGLNet integrating recognition of crop rows and prediction of the effective area of the navigation line is designed. The linear equation of the navigation line we required is obtained through post-processing operations of linear fitting completed in the effective area. According to experimental results, directly teaching the neural network to learn navigation line localization features utilizes the semantic and spatial information in the feature map to establish a stronger relationship between crop rows and the navigation line, which greatly improves the generalization of the navigation line localization in the case of large-angle yaw scenarios without additional training datasets. In addition, the performance of the network is significantly enhanced by adding an attention mechanism in the backbone based on U-Net. Subsequently, a dataset of cabbage vegetable images covering a wide range of complexities in the greenhouse scenario is created, and is available to support future academic research now. Further, a path tracking control scheme based on point-feature and line-feature is proposed, which can control the agricultural vehicle using the information from camera images without the need to perform explicit localization or maintain a global map. Finally, extensive field experiments have been carried out to show that this autonomous navigation system can preserve the stability of the vehicle and ensure timely convergence of deviations in different complex scenarios, which indicates that this control scheme is reliable and robust. Meanwhile, benefit from the efficiency of this recognition algorithm and the unnecessity of the complete representation of the path in the control scheme, the system has high real-time performance in every scenario.

Detection and Tracking of Low-Frame-Rate Water Surface Dynamic Multi-Target Based on the YOLOv7-DeepSORT Fusion Algorithm

This study aims to address the problem in tracking technology in which targeted cruising ships or submarines sailing near the water surface are tracked at low frame rates or with some frames missing in the video image, so that the tracked targets have a large gap between frames, leading to a decrease in tracking accuracy and inefficiency. Thus, in this study, we proposed a water surface dynamic multi-target tracking algorithm based on the fusion of YOLOv7 and DeepSORT. The algorithm first introduces the super-resolution reconstruction network. The network can eliminate the interference of clouds and waves in images to improve the quality of tracking target images and clarify the target characteristics in the image. Then, the shuffle attention module is introduced into YOLOv7 to enhance the feature extraction ability of the target features in the recognition network. Finally, Euclidean distance matching is introduced into the cascade matching of the DeepSORT algorithm to replace the distance matching of IOU to improve the target tracking accuracy. Simulation results showed that the algorithm proposed in this study has a good tracking effect, with an improvement of 9.4% in the improved YOLOv7 model relative to the mAP50-95 value and an improvement of 13.1% in the tracking accuracy in the DeepSORT tracking network compared with the SORT tracking accuracy.

Stand-off concealed firearm detection using motion tracking and convolutional neural networks

<p>The standoff detection of concealed firearms is crutial in managing public security in public spaces. Currently employed standoff concealed weapon detection techniques employ electromagnetic wave imaging which has been found to be extremely slow and may require expensive hardware and may not be applicable in public open spaces. Inorder to maintain safety in open spaces, artificial intelligence enabled video surveillance systems have been widely adopted. This poses an opportunity to explore video surveillance cameras as concealed weapon detectors. A review of existing video surveillance based automated weapon detection approaches discovered that the focus was on the detection of unconcealed firearms leaving a gap in the detection of concealed firearms. This study addresses the aforementioned gap by providing a standoff concealed firearm detection approach on video based on skeletal-based human motion tracking and convolutional neural networks. The motion of armed and unarmed persons was tracked using a depth camera and further classified using convolutional neural networks model. The developed model reported 100% accuracy, precision and recall scores. These results outperformed results obtained from traditional machine learning models therefore highlighting the superior capability of the proposed approach for concealed firearm detection on video to complement the efforts of human video surveillance operators.</p>

Real-Time Bayesian Neural Networks for 6G Cooperative Positioning and Tracking

Real-Time Bayesian Neural Networks for 6G Cooperative Positioning and Tracking

Spatial-Temporal Contextual Aggregation Siamese Network for UAV Tracking

In recent years, many studies have used Siamese networks (SNs) for UAV tracking. However, there are two problems with SNs for UAV tracking. Firstly, the information sources of the SNs are the invariable template patch and the current search frame. The static template information lacks the perception of dynamic feature information flow, and the shallow feature extraction and linear sequential mapping severely limit the mining of feature expressiveness. This makes it difficult for many existing SNs to cope with the challenges of UAV tracking, such as scale variation and viewpoint change caused by the change in height and angle of the UAV, and the challenges of background clutter and occlusion caused by complex aviation backgrounds. Secondly, the SNs trackers for UAV tracking still struggle with extracting lightweight and effective features. A tracker with a heavy-weighted backbone is not welcome due to the limited computing power of the UAV platform. Therefore, we propose a lightweight spatial-temporal contextual Siamese tracking system for UAV tracking (SiamST). The proposed SiamST improves the UAV tracking performance by augmenting the horizontal spatial information and introducing vertical temporal information to the Siamese network. Specifically, a high-order multiscale spatial module is designed to extract multiscale remote high-order spatial information, and a temporal template transformer introduces temporal contextual information for dynamic template updating. The evaluation and contrast results of the proposed SiamST with many state-of-the-art trackers on three UAV benchmarks show that the proposed SiamST is efficient and lightweight.

Siamese Tracking Network with Multi-attention Mechanism

Object trackers based on Siamese networks view tracking as a similarity-matching process. However, the correlation operation operates as a local linear matching process, limiting the tracker’s ability to capture the intricate nonlinear relationship between the template and search region branches. Moreover, most trackers don’t update the template and often use the first frame of an image as the initial template, which will easily lead to poor tracking performance of the algorithm when facing instances of deformation, scale variation, and occlusion of the tracking target. To this end, we propose a Simases tracking network with a multi-attention mechanism, including a template branch and a search branch. To adapt to changes in target appearance, we integrate dynamic templates and multi-attention mechanisms in the template branch to obtain more effective feature representation by fusing the features of initial templates and dynamic templates. To enhance the robustness of the tracking model, we utilize a multi-attention mechanism in the search branch that shares weights with the template branch to obtain multi-scale feature representation by fusing search region features at different scales. In addition, we design a lightweight and simple feature fusion mechanism, in which the Transformer encoder structure is utilized to fuse the information of the template area and search area, and the dynamic template is updated online based on confidence. Experimental results on publicly tracking datasets show that the proposed method achieves competitive results compared to several state-of-the-art trackers.

A method for estimating the rotation errors of BDS broadcast ephemeris using global satellite laser ranging data

Abstract The BeiDou-3 Navigation Satellite System (BDS-3) generates navigation messages from a regional ground tracking network and inter-satellite links, relying on predicted station coordinates and Earth rotation parameters (ERPs). These procedures, however, may lead to orientation errors in the broadcast ephemeris, termed constellation rotation errors. Traditionally, these errors are monitored through Helmert transformation analysis comparing broadcast and post-processed orbits, a method that highlights discrepancies between the global navigation satellite system (GNSS) terrestrial reference frame (TRF) and the International TRF (ITRF) but suffers from a two-week evaluation delay. This study proposes a near real-time approach for estimating constellation rotation errors using high-precision satellite laser ranging (SLR) data, representing the difference between GNSS TRF and ITRF through the International laser ranging service TRF. The impact of SLR data quantity and distribution on rotation estimation accuracy is investigated, and the effectiveness of this method on BDS-3 orbits with ERP-induced orientation deficiencies is validated. Results show that the root mean square of broadcast orbit errors and the orbit user ranging error reduce from 0.8/0.7/0.2 m to 0.1/0.1/0.1 m, respectively. This demonstrates the method’s capability for real-time monitoring and correcting orientation errors within BDS-3 navigation messages.

A varying-parameter complementary neural network for multi-robot tracking and formation via model predictive control

In this paper, a varying-parameter complementary neural network (VPCNN) is designed and combined with model predictive control (MPC) to solve the multi-robot tracking and formation problems via a leader–follower strategy. First, multi-robot tracking and formation problems are transformed into quadratic programming (QP) problems employing an MPC approach. Second, a nonlinear complementary function approach, i.e., the Fischer–Burmeister function, is used to map the Karush–Kuhn–Tucker conditions of the QP problem with double-ended inequality constraints to a system of nonlinear equations. Finally, the VPCNN is designed to solve the multi-robot tracking and formation problem. The effectiveness of the proposed method is demonstrated by numerical simulations, and the advantage of VPCNN in terms of solution speed is indicated by comparisons with a primal–dual neural network.