Tracking Approach Research Articles

Learning Cross-Attention Discriminators via Alternating Time-Space Transformers for Visual Tracking.

In the past few years, visual tracking methods with convolution neural networks (CNNs) have gained great popularity and success. However, the convolution operation of CNNs struggles to relate spatially distant information, which limits the discriminative power of trackers. Very recently, several Transformer-assisted tracking approaches have emerged to alleviate the above issue by combining CNNs with Transformers to enhance the feature representation. In contrast to the methods mentioned above, this article explores a pure Transformer-based model with a novel semi-Siamese architecture. Both the time-space self-attention module used to construct the feature extraction backbone and the cross-attention discriminator used to estimate the response map solely leverage attention without convolution. Inspired by the recent vision transformers (ViTs), we propose the multistage alternating time-space Transformers (ATSTs) to learn robust feature representation. Specifically, temporal and spatial tokens at each stage are alternately extracted and encoded by separate Transformers. Subsequently, a cross-attention discriminator is proposed to directly generate response maps of the search region without additional prediction heads or correlation filters. Experimental results show that our ATST-based model attains favorable results against state-of-the-art convolutional trackers. Moreover, it shows comparable performance with recent "CNN + Transformer" trackers on various benchmarks while our ATST requires significantly less training data.

Robust autonomous detection and tracking of moving objects using hybrid tracking approach

Robust autonomous detection and tracking of moving objects using hybrid tracking approach

Comparing acoustic and satellite telemetry: an analysis quantifying the space use of Chelonia mydas in Bimini, Bahamas.

Passive acoustic and Argos satellite telemetry are common methods for tracking marine species and are often used similarly to quantify space use. However, data-driven comparisons of these methods and their associated ecological inferences are limited. To address this, we compared temporal durations, spatial resolutions, financial costs and estimates of occurrence and range distributions for each tracking approach using nine juvenile green turtles (Chelonia mydas) in Bimini, Bahamas. Tracking durations were similar, although acoustic tracking provided higher spatiotemporal resolution than satellite tracking. Occurrence distributions (95%) estimated from satellite telemetry were 12 times larger than those from acoustic telemetry, while satellite range distributions (95%) were 89 times larger. While individuals generally remained within the extent of the acoustic receiver array, gaps in coverage were identified. These gaps, combined with the lower accuracy of satellite telemetry, were likely drivers for the larger satellite distributions. Costs differed between telemetry methods, with acoustic telemetry being less expensive at larger sample sizes with a previously established array. Our results suggest that acoustic and satellite telemetry may not provide similar inferences of individual space use. As such, we provide recommendations to identify telemetry methods appropriate for specific study objectives and provide discussion on the biases of each.

A Novel Planning and Tracking Approach for Mobile Robotic Arm in Obstacle Environment

In this study, a novel planning and tracking approach is proposed for a mobile robotic arm to grab objects in an obstacle environment. First, we developed an improved APF-RRT* algorithm for the motion planning of a mobile robotic arm. This algorithm optimizes the selection of random tree nodes and smoothing the path. The invalid branch and the planning time are decreased by the artificial potential field, which is determined by the specific characteristics of obstacles. Second, a Fuzzy-DDPG-PID controller is established for the mobile robotic arm to track the planned path. The parameters of the PID controller are set using the new DDPG algorithm, which integrated FNN. The response speed and control accuracy of the controller are enhanced. The error and time of tracking of the mobile robotic arm are decreased. The experiment results verify that the proposed approach has good planning and tracking results, high speed and accuracy, and strong robustness. To avoid the occasionality of the experiments and fully illustrate the effectiveness and generality of the proposed approach, the experiments are repeated multiple times. The experiment results demonstrate the effectiveness of the proposed approach. It outperforms existing planning and tracking approaches.

Scalable multitarget tracking using PCL in SFN with 3D data association uncertainty

Passive coherent location (PCL) has attracted widespread attention due to its low cost, good concealment, and strong anti-stealth. When multiple illuminators operating in a single frequency network are used in a PCL system for multitarget tracking, the three-dimensional (3D) data association uncertainty among the illuminators, measurements and targets has to be dealt with because it will significantly increase computational complexity and reduce scalability. To solve this problem, a scalable multi-association belief propagation (MA-BP) approach for multitarget tracking is proposed. By factorizing the joint probability density function, the tracking problem is described by a factor graph, in which two new illuminator-target oriented and measurement oriented association variables are defined to describe the 3D data association uncertainty. Furthermore, under the Gaussian assumption, a Gaussian version of MA-BP, named as multi-association Gaussian belief propagation (MA-GaBP), is implemented with a generalized virtual measurement model. The MA-GaBP approach can further reduce memory requirements and computational complexity. Simulation results show that the proposed approach effectively solves the 3D data association uncertainty, and has satisfied tracking accuracy and scalability.

DEVELOPMENT OF ARTS LEARNING MEDIA USING AUGMENTED REALITY WITH MARKERLESS BASED TRACKING METHOD AT SMKN 47

During the transitional time between Covid-19 and endemic, all Indonesian schools, and SMKN 47 in particular, had restricted educational opportunities. Due to challenges encountered by the instructor and 36 students majoring in accounting in grade 12, they were unable to carry out teaching and learning activities at school; art lesson formats with three-dimensional learning modules must be visualized in 3D. However, with hybrid teaching and learning utilizing presentations, displaying the learned 3D forms is less successful. In order to visualize three dimensions in fine arts classes using augmented reality technology, we require a new learning medium. The application of the Markerless Based Tracking approach in this work enables the presentation of a tracked 3D model in the surrounding environment in real-time by merging the actual and virtual worlds as if their boundaries did not exist. The augmented reality system scans flat surfaces utilizing points, as opposed to markers or other auxiliary media. Points used as pedestals or containers to elevate three-dimensional items. There are three sculptures and three traditional dwellings on show. Online examination outcomes average resulted a 78.2% that the AR Fine Arts application by instructors and students are therefore consistent and well accepted.

Vision-based autonomous navigation stack for tractors operating in peach orchards

Recently, camera-based approaches have been proposed to replace expensive LiDAR or real-time kinematic (RTK) based solutions for the unmanned tractors operating in an orchard-like environment. However, constructing a completely autonomous navigation stack with a low-costly camera and onboard computer presents extra difficulties, such as a computationally efficient perception model, rapid path planning without costly SLAM, and the capacity to avoid obstacles. In this paper, a novel vision-based autonomous navigation stack is devised to address challenges of the wholly autonomous tractor using stereo camera and inertial measurement unit (IMU) that are inexpensive. The computational pipeline consists primarily of three modules: (1) the multi-task perception network, (2) the frame transformation algorithm, and (3) the motion planning module. The multi-task perception network detects tree trunks, obstacles, and traversable areas simultaneously from the RGB image with high efficiency (69 FPS) and accuracy (mAP@.5 of 96.7% and mIoU of 98.1%). For global path planning and trajectory prediction, the frame transformation algorithm integrates the downsized the navigational features and transforms them to the tractor frame from the image frame. The motion planning module then uses the fused data to plan the appropriate path (i.e., the center of the tree row, U-turn path, or J-turn path) and search for the optimal trajectory utilizing the optimized dynamic window approach (DWA) for path tracking. In the peach orchard, the proposed stack is implemented on our autonomous retrofitted tractor and its efficacy is demonstrated compared with the human-driving tractor.

Logistics Management System using Big Data Analytics

Nowadays, the e-commerce sector has emerged as a significant player in the current financial landscape. It offers a wide range of advantages for global customers, and given the increasing demand for various services and products, there is a pressing need for the development of efficient Web applications. This e-commerce application serves as the bridge between small businesses and their clientele. The application harnesses the capabilities of various technologies, including MongoDB for data storage, Node.js and Express.js for a fast and efficient backend, and HTML CSS for constructing a user-friendly interface for clients. This thesis underscores the proficiency of the (MongoDB, Express.js, HTML, CSS, Node.js) technology stack in the creation of a comprehensive e-commerce application with robust scalability options. The study delves into a meticulous development process, spanning from the backend to the frontend. The results unequivocally underscore the indispensable role in crafting a customizable Web application. We have also investigated the realm of big data analytics and its applications in logistics and supply chain management. It delves into innovative methods, practices, and prospects within this field. The articles within this collection scrutinize multiple avenues for enhancing big data analytics and its applications in logistics and supply chain management. These include the exploration of technology-centric tracking approaches, the examination of the correlation between financial performance and data-driven supply chains, as well as the assessment of implementation challenges and the maturity of supply chain capabilities in the context of big data.

Multiple voltage source converters based microgrid with solar photovoltaic array and battery storage

In this paper, a standalone photovoltaic (PV)-battery storage (BS) based microgrid (MG) is presented with a 415V-AC bus. The PV array is linked to the main voltage source converter (VSC1) DC-link to maximize power production in an efficient and cost-effective single-stage structure. This is achieved through an incremental conductance (INC) maximum power point (MPP) tracking approach. Moreover, a secondary converter (VSC2) is connected to BS for power flow management. MG performance is evaluated under various operating conditions. Steady-state and dynamic performances of MG are studied for variations such as changes in solar PV irradiance and load unbalances.

An artificial neural network for full-body posture prediction in dynamic lifting activities and effects of its prediction errors on model-estimated spinal loads

Musculoskeletal models have indispensable applications in occupational risk assessment/management and clinical treatment/rehabilitation programs. To estimate muscle forces and joint loads, these models require body posture during the activity under consideration. Posture is usually measured via video-camera motion tracking approaches that are time-consuming, costly, and/or limited to laboratories. Alternatively, posture-prediction tools based on artificial intelligence can be trained using measured postures of several subjects performing many activities. We aimed to use our previous posture-prediction artificial neural network (ANN), developed based on many measured static postures, to predict posture during dynamic lifting activities. Moreover, effects of the ANN posture-prediction errors on dynamic spinal loads were investigated using subject-specific musculoskeletal models. Seven individuals each performed twenty-five lifting tasks while their full-body three-dimensional posture was measured by a 10-camera Vicon system and also predicted by the ANN as functions of the hand-load positions during the lifting activities. The measured and predicted postures (i.e., coordinates of 39 skin markers) and their model-estimated L5-S1 loads were compared. The overall root-mean-squared-error (RMSE) and normalized (by the range of measured values) RMSE (nRMSE) between the predicted and measured postures for all markers/tasks/subjects was equal to 7.4 cm and 4.1 %, respectively (R2 = 0.98 and p < 0.05). The model-estimated L5-S1 loads based on the predicted and measured postures were generally in close agreements as also confirmed by the Bland-Altman analyses; the nRMSE for all subjects/tasks was < 10 % (R2 > 0.7 and p > 0.05). In conclusion, the easy-to-use ANN can accurately predict posture in dynamic lifting activities and its predicted posture can drive musculoskeletal models.

Multiple maneuvering target tracking based on hierarchical Dirichlet process and hidden Markov model

This paper is concerned with multiple maneuvering target tracking problems. Although the interacting multiple model (IMM) algorithm can be used for maneuvering targets tracking, the performance is limited as the motion mode set of the target is fixed in advance. In this paper, we address the problem of multiple maneuvering targets with unknown maneuvering models in cluttered environments. A finite hidden Markov model (HMM) is developed with truncated Hierarchical Dirichlet process (HDP) to recognize the unknown and changing motion modes. Further, the HDP-HMM is introduced into multiple hypothesis tracking (MHT) and probabilistic multiple hypothesis tracking (PMHT) approaches for multiple maneuvering targets tracking with data association uncertainty. To joint estimate the HDP-HMM parameters and target states, the particle learning and variational inference approaches are used, and the solutions of HDP-HMM-MHT and HDP-HMM-PMHT algorithms are obtained based on Rao-Blackwellized particle filter and variational Bayesian methods, respectively. As the changes of maneuver models can be learned using HDP, the multi-target tracking accuracy of the proposed algorithm is improved. Simulation results show that the proposed algorithms outperform the traditional IMM-MHT method with fixed model sets.

MIPS-Fusion: Multi-Implicit-Submaps for Scalable and Robust Online Neural RGB-D Reconstruction

We introduce MIPS-Fusion, a robust and scalable online RGB-D reconstruction method based on a novel neural implicit representation - multi-implicit-submap. Different from existing neural RGB-D reconstruction methods lacking either flexibility with a single neural map or scalability due to extra storage of feature grids, we propose a pure neural representation tackling both difficulties with a divide-and-conquer design. In our method, neural submaps are incrementally allocated alongside the scanning trajectory and efficiently learned with local neural bundle adjustments. The submaps can be refined individually in a back-end optimization and optimized jointly to realize submap-level loop closure. Meanwhile, we propose a hybrid tracking approach combining randomized and gradient-based pose optimizations. For the first time, randomized optimization is made possible in neural tracking with several key designs to the learning process, enabling efficient and robust tracking even under fast camera motions. The extensive evaluation demonstrates that our method attains higher reconstruction quality than the state of the arts for large-scale scenes and under fast camera motions.

Backstepping Controller for Mobile Robot in Presence of Disturbances and Uncertainties

The objective of this work is to devise an effective control system for addressing the trajectory tracking challenge in nonholonomic mobile robots. Two primary control approaches, namely kinematic and dynamic strategies, are explored to achieve this goal. In the kinematic control domain, a backstepping controller (BSC) is introduced as the core element of the control system. The BSC is utilized to guide the mobile robot along the desired trajectory, leveraging the robot’s kinematic model. To address the limitations of the kinematic control approach, a dynamic control strategy is proposed, incorporating the dynamic parameters of the robot. This dynamic control ensures real-time control of the mobile robot. To ensure the stability of the control system, the Lyapunov stability theory is employed, providing a rigorous framework for analyzing and proving stability. Additionally, to optimize the performance of the control system, a genetic algorithm is employed to design an optimal control law. The effectiveness of the developed control approach is demonstrated through simulation results. These results showcase the enhanced performance and efficiency achieved by the proposed control strategies. Overall, this study presents a comprehensive and robust approach for trajectory tracking in nonholonomic mobile robots, combining kinematic and dynamic control strategies while ensuring stability and performance optimization.

Reactive mobile manipulation based on dynamic dual-trajectory tracking

The letter presents an online obstacle avoidance strategy for a nonholonomic mobile manipulator, including dual-trajectory generation for the mobile base(MB) and end effector(EE), as well as whole-body reactive obstacle avoidance. The feasible region of the manipulator’s workspace is generated by traversing the path of the MB, and the spatio-temporal trajectory of the EE is generated by quadratically constrained quadratic programming(QCQP) planning, satisfying the physical constraints and obstacle avoidance constraints. The reactive obstacle avoidance strategy is based on quadratic programming(QP), using singularity avoidance and EE trajectory tracking tasks as soft constraints with different priorities. Considering the nonholonomic constraints and motion differences of MB, we employ a dynamic tracking approach to avoid obstacles of MB. Additionally, an adaptive weight method is utilized to adjust the priority of MB path tracking task, effectively preventing joint velocity jitter during the insertion process. Our strategy has been proven to generate obstacle avoidance trajectories for EE in real-time and ensure safe joint velocities for MB and the manipulator through simulation and real-world scenarios.

Deep Open-Curve Snake for Discriminative 3D Neuron Tracking.

Open-Curve Snake (OCS) has been successfully used in three-dimensional tracking of neurites. However, it is limited when dealing with noise-contaminated weak filament signals in real-world applications. In addition, its tracking results are highly sensitive to initial seeds and depend only on image gradient-derived forces. To address these issues and boost the canonical OCS tracker to a new level of learnable deep learning algorithms, we present Deep Open-Curve Snake (DOCS), a novel discriminative 3D neuron tracking framework that simultaneously learns a 3D distance-regression discriminator and a 3D deeply-learned tracker under the energy minimization, which can promote each other. In particular, the open curve tracking process in DOCS is formed as convolutional neural network prediction procedures of new deformation fields, stretching directions, and local radii and iteratively updated by minimizing a tractable energy function containing fitting forces and curve length. By sharing the same deep learning architectures in an end-to-end trainable framework, DOCS is able to fully grasp the information available in the volumetric neuronal data to address segmentation, tracing, and reconstruction of complete neuron structures in the wild. We demonstrated the superiority of DOCS by evaluating it on both the BigNeuron and Diadem datasets where consistently state-of-the-art performances were achieved for comparison against current neuron tracing and tracking approaches. Our method improves the average overlap score and distance score about 1.7% and 17% in the BigNeuron challenge data set, respectively, and the average overlap score about 4.1% in the Diadem dataset.

Developing Topics.

Alzheimer's disease is characterized by cortical hyperexcitability in the early stages, which may be attributed to the intricate interplay between tau and amyloid β pathologies. To combat the propagation of tau pathology and cognitive decline, it may be beneficial to maintain a high degree of functional brain network segregation. Additionally, the use of non-invasive brain stimulation techniques to alter cortical excitability could help alleviate cognitive deficits. However, it is unclear how these interventions are related to tau or amyloid biomarkers. In this study, 530 participants aged 40 to 65 from the Barcelona Brain Health Initiative cohort were analyzed to investigate the connection between plasma concentration of tau phosphorylated at amino acid 181 (pTau181), cortical excitability, and segregation of functional brain networks. Participants had functional magnetic resonance imaging (fMRI), electroencephalography (EEG), and pTau181 available. A subset of 47 participants also underwent transcranial magnetic stimulation with concurrent EEG (TMS-EEG). Cortical excitability was assessed using the one-over-f slope of the EEG power spectrum (1/f-slope) and evoked response to a TMS perturbation (TEP) of the left prefrontal cortex (L-PFC) between 160-240ms post single-pulse TMS. The degree of segregation between major functional networks was measured from fMRI data using the system segregation statistic. A generalized linear model with a gamma distribution (Table 1) revealed that steeper 1/f-slope correlated with higher pTau181 concentration, particularly in participants older than 53 years of age (Figure 1.A), or system segregation below 0.28 (Figure 1.B). A multiple linear regression within the TMS-EEG subsample (Table 2) revealed that the TEP response correlated positively with pTau181 as age increased, and that this model, despite including a smaller sample size, better explained pTau181 concentrations. The study shows that cortical excitability is related to pTau181 based on age and system segregation. The TEP response is a more sensitive marker than unperturbed EEG metrics in explaining the relation between excitability and pTau181 concentrations. Combined TMS with EEG could potentially be an inexpensive and scalable approach for early identification and longitudinal tracking of middle-aged adults at risk for cognitive decline and dementia.

Poster Session I: LissEYEjous Tracker - precise fundus tracking device based on ultrafast Lissajous scanning.

Retinal eye tracking has emerged as a promising alternative to conventional video-based trackers, offering direct access to retinal coordinates with refined spatial and temporal resolutions. These attributes make them attractive for applications ranging from image stabilization in advanced ophthalmic imaging to identifying biomarkers of neurological or ophthalmic disorders that affect eye motility. Existing retinal tracking method however face challenges related to reliance on reference frames and non-uniform sampling either in space or time. In this work we present a new approach for retinal tracking, which is based on imaging small retinal patches (~1.5-3°) using self-repeating Lissajous scanning patterns. Pattern repetition rates close to 4kHz are achieved with an optical design that employs two MEMS microscanners with closely matching resonant frequencies, working in mutually perpendicular dimensions. Several examples of fundus images acquired with different Lissajous patterns are presented. Based on this, eye trajectories may be extracted. Future works will further investigate tracking resolution and dependence on Lissajous pattern spatial density and repetition rate.

A Novel Approach for Train Tracking in Virtual Coupling Based on Soft Actor-Critic

The development of virtual coupling technology provides solutions to the challenges faced by urban rail transit systems. Train tracking control is a crucial component in the operation of virtual coupling, which plays a pivotal role in ensuring the safe and efficient movement of trains within the train and along the rail network. In order to ensure the high efficiency and safety of train tracking control in virtual coupling, this paper proposes an optimization algorithm based on Soft Actor-Critic for train tracking control in virtual coupling. Firstly, we construct the train tracking model under the reinforcement learning architecture using the operation states of the train, Proportional Integral Derivative (PID) controller output, and train tracking spacing and speed difference as elements of reinforcement learning. The train tracking control reward function is designed. Then, the Soft Actor-Critic (SAC) algorithm is used to train the virtual coupling train tracking reinforcement learning model. Finally, we took the Deep Deterministic Policy Gradient as the comparison algorithm to verify the superiority of the algorithm proposed in this paper.

Machine learning based cascaded ANN MPPT controller for erratic PV shading circumstances

Power generation is challenged to meet energy demand during peak hours. As a result of limited non-renewable energy resources, power utilities heavily rely on fossil fuels. Therefore, scientists and researchers are looking for some distributed generators to provide additional power during peak hours. During such period, load demand is solved using solar power. As a consequence, grid-connected solar Photovoltaic (PV) systems are catching the attention owing to their ability to significantly reduce the use of fossil fuels. Under Partial Shading Condition (PSC), this paper utilizes Luo Converter along with Cascaded Artificial Neural Network (ANN), which is a Machine Learning Based Maximum Power Point Tracking (ML-MPPT) approach for tracking optimal power from PV system. The gained DC supply is converted into AC voltage using VSI attached to the system. In addition, PI controller engaged controls the voltage at grid side and results in effective grid synchronization. Furthermore, MATLAB Simulink analysis is carried out and the outcomes reveal the effectiveness of proposed system with 98% efficiency under different PV circumstances.

A Value Modeling Approach for Product Development Processes and Supply Chain Management

This research presents a quantifiable value modeling approach for tracking and optimizing value addition at different stages of the product development process (PDP) and supply chain management (SCM). While there is a vast abundance of papers in the literature harping on the need to maximize the value, especially in lean thinking context, in PDP and SCM, there is a dearth of propositions on how to quantitatively measure and optimize value at each stage of product development and supply chain. This paper presents a multi-dimensional value modeling approach that incorporates parameters and variables from different stages of PDP and SCM by mathematically mapping them on to two macroscopic metrics, namely Product Information Evolution and Risk Reduction. These two metrics are in turn modeled by incorporating value attribute variables such as performance, risk, schedule, cost, form, fit, function and timeliness from different stages of PDP-SCM such as customer needs ideation, product functions, inventory management, logistics management and product delivery. The efficacy of the proposed model was tested by utilizing it to measure value additions at different stages of PDP-SCM as well as the total value addition for an ordinance disposal (OD) robot product family, once developed and supplied using traditional production method and once using lean production method, with the underlying axiom being the lean method will result in higher value. To ensure a fair and accurate comparison, both production methods used the proposed mathematical value model and worked toward maximizing total value using nonlinear programming. The results showed that lean PDP-SCM method added higher value at each of the stages as well as total value when compared to the traditional PDP-SCM method thus establishing the proposed value model as a viable mathematical way to measure value. While the proposed value model passed test of efficacy against the lean axiom, further research is needed to test it against other established axioms to establish it as a robust mathematical way to measure value in PDP-SCM.