Top Tracker Research Articles

MIMTracking: Masked image modeling enhanced vision transformer for visual object tracking

Recently, one-stage trackers achieve state-of-the-art tracking results due to more sufficient integration of search and template representations. These methods usually adopt an encoder for synchronous feature generation and interaction. Despite their high performance, the one-stage approaches tend to feed the encoder full input representations that are highly redundant and dense during training. With a focus on tackling this problem, a novel algorithm termed MIMTracking is developed for visual target tracking. MIMTracking exploits an encoder and a decoder for masked image modeling during training. Randomly sampled discrete search embeddings and template embeddings serve as input of the encoder. The lightweight decoder takes full representations as input and progressively highlights the target region. This design alleviates input redundancy and reduces the computational cost of the training process, thereby allowing for more efficient learning of useful representations. The proposed MIMTracking achieves state-of-the-art tracking results on numerous tracking datasets, e.g., 51.5 % area under curve (AUC) on LaSOText, outperforming the previous top tracker OSTrack by 2 %. Especially, our large model MIMTracking-L further improves the AUC to 53.4 % on LaSOText.

The JUNO experiment Top Tracker

The main task of the Top Tracker detector of the neutrino reactor experiment Jiangmen Underground Neutrino Observatory (JUNO) is to reconstruct and extrapolate atmospheric muon tracks down to the central detector. This muon tracker will help to evaluate the contribution of the cosmogenic background to the signal. The Top Tracker is located above JUNO’s water Cherenkov Detector and Central Detector, covering about 60% of the surface above them. The JUNO Top Tracker is constituted by the decommissioned OPERA experiment Target Tracker modules. The technology used consists in walls of two planes of plastic scintillator strips, one per transverse direction. Wavelength shifting fibres collect the light signal emitted by the scintillator strips and guide it to both ends where it is read by multianode photomultiplier tubes. Compared to the OPERA Target Tracker, the JUNO Top Tracker uses new electronics able to cope with the high rate produced by the high rock radioactivity compared to the one in Gran Sasso underground laboratory. This paper will present the new electronics and mechanical structure developed for the Top Tracker of JUNO along with its expected performance based on the current detector simulation.

Deferred Optical Photon simulation for the JUNO experiment

The Jiangmen Underground Neutrino Observatory (JUNO) is designed to determine the neutrino mass ordering and precisely measure oscillation parameters. It is being built in South China at a depth of 700 m underground and comprises a central detector, water Cerenkov detector and top tracker. The central detector is designed to detect anti-neutrinos with an energy resolution of 3% at 1 MeV, using a 20 kt liquid scintillator target with 17,612 20-inch PMTs and 25,600 3-inch PMTs. The scintillator provides a light yield of approximately 10,000 photons per MeV. Monte Carlo simulation is a crucial tool for developing an understanding of detector performance, requiring the production of large samples of background processes with optical photons. Simulation of large numbers of optical photons with Geant4 is computationally challenging for both processing time and memory resources. In order to optimize resource usage, a deferred optical photon simulation workflow is proposed and implemented using Geant4 classes. The key idea is to simulate events initially without optical photons, only performing the optical photon simulation when user specified criteria are met.In this contribution, the design and the implementation of the deferred optical photon simulation will be presented. Optical simulation comprises generation of photons and propagation through the detector implementing optical physics processes including reflection, refraction, scattering and absorption. Instead of generating the optical photons at each step by Geant4 immediately, the necessary data to generate optical photons at each step are collected, which is termed GenStep. At the end of each event, user specified criteria determines if the optical photon simulation is performed using a class called G4OPSimulator. The class G4OPSimulator implements a customized simulation workflow, based on Geant4 internal classes including G4TrackingManager and G4StackManager. The simulator passes references to the collected GenStep objects to customized Scintillation and Cerenkov processes which generate the optical photons in G4Track objects. As a track could be absorbed and re-emitted, the secondaries will be retrieved from the G4TrackingManager and pushed to the G4StackManager. The performance of the simulator will be presented. The technique of deferred optical photon simulation can be applied to all event types and it is expected to be particularly beneficial with rare processes. Especially the events must be selected during the detector simulation at runtime, instead of the event generation. An application of the technique to the simulation of such events will be shown.

Muon reconstruction with a convolutional neural network in the JUNO detector

The Jiangmen Underground Neutrino Observatory (JUNO) is designed to determine the neutrino mass ordering and measure neutrino oscillation parameters. A precise muon reconstruction is crucial to reduce one of the major backgrounds induced by cosmic muons. This article proposes a novel muon reconstruction method based on convolutional neural network (CNN) models. In this method, the track information reconstructed by the top tracker is used for network training. The training dataset is augmented by applying a rotation to muon tracks to compensate for the limited angular coverage of the top tracker. The muon reconstruction with the CNN model can produce unbiased tracks with performance that spatial resolution is better than 10 cm and angular resolution is better than 0.6 degrees. By using a GPU accelerated implementation a speedup factor of 100 compared to existing CPU techniques has been demonstrated.

Feature Aggregation Networks Based on Dual Attention Capsules for Visual Object Tracking

Tracking-by-detection algorithms have considerably enhanced tracking performance with the introduction of recent convolutional neural networks (CNNs). However, most trackers directly exploit standard scalar-output CNN features, which may not capture enough feature encoding information, instead of aggregated CNN features of vector-output form. In this paper, we propose an end-to-end feature aggregation capsule framework. First, based on the existing CNN network, we aggregate a certain number of similar position-aware CNN features into a capsule to model the feature similarity. The acquired vector-level feature capsules (rather than previous scalar-level pointwise features) are utilized for differentiation learning. We then propose a group attention module to better model the entity representation between different capsule groups thus optimizes total discriminative capability. Third, to reduce the prediction interference resulted by the side effect of dimension rising within capsules, we propose a penalty attention module. Such strategy could dynamically adjust values of neurons by estimating whether they are beneficial or harmful to tracking. Experimental results on five representative benchmarks (UAVDT, DTB70, UAV123, VOT2016 and VOT2018) demonstrate the excellent tracking performance of our dual attention based capsule tracker (DACapT). Specially, it exceeds the previous top tracker by 4.6%/1.9% in precision/success evaluations on UAVDT.

Adaptive Multi-Feature Reliability Re-Determinative Correlation Filter for Visual Tracking

Model drift is a challenging issue for visual object tracking. Most available approaches aim to address this issue by constructing of a stronger discriminative prediction model with the integration of multiple features. In this article, we also address the issue of model drift using multiple features, but we consider the fusion problem from a different point of view, namely, strengthening the features that are suitable for the current scenario while weakening the remaining ones. Therefore, an advanced tracker that adaptively redetermines the reliability for each feature during the tracking process is proposed. Furthermore, to correctly evaluate and redetermine these reliabilities, two different solutions, called model evaluation and numerical optimization, are proposed, and two independent trackers corresponding to these two solutions are implemented. Extensive experiments have been designed on five large datasets to validate the following: 1) The proposed tracking framework is superior for making the tracking model more robust, and 2) the two solutions proposed for redetermining the reliability for each feature are effective. As expected, the two implemented trackers do indeed improve the accuracy and robustness compared to state-of-the-art trackers. Especially on VOT2016, the proposed trackers based on model evaluation and numerical optimization achieve outstanding EAO scores (i.e., <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.453</b> and <bold xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">0.428</b> , respectively), outperforming the recently developed top trackers by a large margin.

In-Depth Evaluation of Redirect Tracking and Link Usage

Abstract In today’s web, information gathering on users’ online behavior takes a major role. Advertisers use different tracking techniques that invade users’ privacy by collecting data on their browsing activities and interests. To preventing this threat, various privacy tools are available that try to block third-party elements. However, there exist various tracking techniques that are not covered by those tools, such as redirect link tracking. Here, tracking is hidden in ordinary website links pointing to further content. By clicking those links, or by automatic URL redirects, the user is being redirected through a chain of potential tracking servers not visible to the user. In this scenario, the tracker collects valuable data about the content, topic, or user interests of the website. Additionally, the tracker sets not only thirdparty but also first-party tracking cookies which are far more difficult to block by browser settings and ad-block tools. Since the user is forced to follow the redirect, tracking is inevitable and a chain of (redirect) tracking servers gain more insights in the users’ behavior. In this work we present the first large scale study on the threat of redirect link tracking. By crawling the Alexa top 50k websites and following up to 34 page links, we recorded traces of HTTP requests from 1.2 million individual visits of websites as well as analyzed 108,435 redirect chains originating from links clicked on those websites. We evaluate the derived redirect network on its tracking ability and demonstrate that top trackers are able to identify the user on the most visited websites. We also show that 11.6% of the scanned websites use one of the top 100 redirectors which are able to store nonblocked first-party tracking cookies on users’ machines even when third-party cookies are disabled. Moreover, we present the effect of various browser cookie settings, resulting in a privacy loss even when using third-party blocking tools.

Progress of Veto Detector of JUNO

The Jiangmen Underground Neutrino Observatory (JUNO) is a 20 kiloton liquid scintillator detector with neutrino mass hierarchy determination as primary physics goal. The other measurements, including precise neutrino oscillation parameters, solar neutrino, geo-neutrino, supernova neutrinos and the diffuse supernova neutrinos background, will be performed at the same time. The detector will be built in a 700 meter deep underground laboratory. A multi-eto detector, composed of a Top Tracker and a Water Cherenkov, will be installed for cosmic muon detection and background reduction. The Top Tracker detector is placed above the central detector, and is composed of three layers of scintillating strips. The outer part of the Central Detector is filled with water and equipped with about 2400 MCP-PMTs (20 inches) to form the Water Cherenkov detector used for muon tagging. Thanks to the Veto detector, the cosmic muon induced fast neutron background can be reduced to a level of about 0.06/day.

Development of the JUNO Conditions Data Management System

The Jiangmen Underground Neutrino Observatory (JUNO) experiment is designed to determine the neutrino mass hierarchy and precisely measure oscillation parameters with an unprecedented energy resolution of 3% at 1 MeV. It is composed of a 20 kton liquid scintillator central detector equipped with 18000 20-inch PMTs and 25000 3-inch PMTs, a water pool with 2000 20-inch PMTs, and a top tracker. Conditions data, coming from calibration and detector monitoring, are heterogeneous, different type of conditions data has different write rates, data format and data volume. JUNO conditions data management system (JCDMS) is developed to homogeneously treat all these heterogeneous conditions data in order to provide easy management and convenient access with both Restful API and web interfaces, support good scalability and maintenance for long time running. The paper describes the status and development of JCDMS including the data model, workflows, interfaces, data caching and performance of the system.

Status of JUNO Simulation Software

The JUNO (Jiangmen Underground Neutrino Observatory) experiment is a multi-purpose neutrino experiment designed to determine the neutrino mass hierarchy and precisely measure oscillation parameters. It will be composed of a 20k ton liquid scintillator (LS) central detector equipped with about 18000 20-inch photon-multipliers (PMTs) and 25000 3-inch PMTs, a water Cherenkov detector with about 2000 20-inch PMTs, and a top tracker. Monte-Carlo simulation is a fundamental tool for optimizing the detector design, tuning reconstruction algorithms, and performing physics study. The status of JUNO simulation software will be presented, including generator interface, detector geometry, physics processes, MC truth, pull-mode electronic simulation.

Patch-based adaptive weighting with segmentation and scale (PAWSS) for visual tracking in surgical video.

Vision-based tracking in an important component for building computer assisted interventions in minimally invasive surgery as it facilitates estimation of motion for instruments and anatomical targets. Tracking-by-detection algorithms are widely used for visual tracking, where the problem is treated as a classification task and a tracking target appearance model is updated over time using online learning. In challenging conditions, like surgical scenes, where tracking targets deform and vary in scale, the update step is prone to include background information in model appearance or to lack the ability to estimate change of scale, which degrades the performance of classifier. In this paper, we propose a Patch-based Adaptive Weighting with Segmentation and Scale (PAWSS) tracking framework that tackles both scale and background problems. A simple but effective colour-based segmentation model is used to suppress background information and multi-scale samples are extracted to enrich the training pool, which allows the tracker to handle both incremental and abrupt scale variations between frames. Experimentally, we evaluate our approach on Online Tracking Benchmark (OTB) dataset and Visual Object Tracking (VOT) challenge datasets, showing that our approach outperforms recent state-of-the-art trackers, and it especially improves successful rate score on OTB dataset, while on VOT datasets, PAWSS ranks among the top trackers while operating at real-time frame rates. Focusing on the application of PAWSS to surgical scenes, we evaluate on MICCAI 2015 challenge instrument tracking challenge and in vivo datasets, showing that our approach performs the best among all submitted methods and also has promising performance on in vivo surgical instrument tracking.

The Jiangmen Underground Neutrino Observatory (JUNO)

The Jiangmen Underground Neutrino Observatory (JUNO) is an experiment under construction in China with the primary goal of determining the neutrino mass hierarchy (MH) with reactor anti-neutrinos. The JUNO detector system consists of a central detector, an active veto system and a calibration system. The central detector is a 35 m diameter transparent acrylic sphere containing a 20 kton liquid scintillator neutrino target. A primary photo-detection system consisting of 18,000 large (20” diameter) dynode and micro-channel plate photomultipliers surrounds the central detector. A second interlaced photo-detection system is made of 25,600 small (3” diameter) photomultipliers working in the single photo-electron regime for the reactor anti-neutrino detection. The detector is designed to achieve an unprecedented energy resolution of 3% @1MeV and an absolute energy scale uncertainty better than 1%. A veto system, consisting of a water Cherenkov detector and a top tracker, is used help maximally remove cosmogenic backgrounds. Due to its unprecedented scale and precision, JUNO will be an exceptional multipurpose detector with a rich physics program in neutrino oscillation, geo-neutrinos, astrophysical neutrinos and the search for physics beyond the Standard Model (sterile neutrinos, dark matter, proton decay and others).

Exploiting weak mask representation with convolutional neural networks for accurate object tracking

Recent years have witnessed the popularity of Convolutional Neural Networks (CNN) in a variety of computer vision tasks, including video object tracking. Existing object tracking methods with CNN employ either a scalar score or a confidence map as CNN’s output, which suffer the infeasibility of estimating the object’s accurate scale and rotation angle. Specifically, as with other traditional methods, they assume the targets’ scale aspect ratio and rotation angle are fixed. To address the limitation, we propose to take a binary mask as the output of CNN for tracking. To this end, we adapt a semantic segmentation model by online fine-tuning with augmented samples in the initial frame to uncover the target in the following frames. During the generation of training samples, we employ a Crop and Paste method to better utilize context information, add a random value to lightness component to mimic the illumination change, and take a Gaussian filtering approach to mimic the blur. During the tracking, due to the limitation of CNN’s receptive field size and spatial resolution, the network may fail to identify the target if the estimated bounding box is considerably incorrect. Therefore we propose a bounding box approximation method by considering temporal consistency. Excluding the initial training cost, our tracker runs at 41 FPS on a single GeForce 1080Ti GPU. Evaluated on benchmarks including OTB-2015, VOT-2016 and TempleColor, it achieves comparable results with non real-time top trackers and state-of-the-art performance among those real-time ones.

The future of the high energy cosmic ray detection: HERD

The High Energy cosmic-Radiation Detection (HERD) facility will be one of the space astronomy payloads on board the future Chinese space station. The ambitious aim of HERD is the direct detection of cosmic rays towards the “knee” region (~ 1 PeV), with a detector able to measure electrons, photons and nuclei with an excellent energy resolution (1% for electrons and photons at 200 GeV and 20% for nuclei at 100 GeV - PeV), an acceptance 10 times the one of present generation missions (~ 1 m2 sr), and long life-time (&gt; 10 years). The primary objectives of HERD are the indirect search for dark matter particles and the precise measurement of energy distribution and composition of cosmic rays from 30 GeV up to a few PeV, determining the origin of the “knee” structure of the spectrum. Furthermore, HERD will monitor the high energy gamma-ray sky from 500 MeV, observing gamma-ray bursts, active galactic nuclei, galactic microquasars, etc. HERD will be composed of a homogeneous calorimeter, surrounded by a particle tracker and a plastic scintillator detector. Two possible trackers are under study: a 5-side tracker made of silicon strip detectors and a 4-side scintillating fiber tracker with a silicon strip top tracker. The total volume of HERD will be (2.3 × 2.3 × 2.6) m3 with a weight of about 4 t. The HERD design, perspectives, expected performances in terms of energy sensitivity and acceptance will be presented in this contribution.

Shielding the Earth Magnetic Field using Spherical Coils

The Jiangmen Underground Neutrino Observatory (JUNO) is a neutrino experiment consisting of 2 systems; central detector (CD) and veto systems. The CD is composed of thousands of photomultiplier tubes (PMTs) used to detect light signals from neutrino induced interactions with liquid scintillator inside the CD. Another set of PMTs is used as water Cherenkov detector, together with muon top tracker forming the veto system for background rejection [1]. However, the PMTs’ efficiency decreases when they are used in magnetic field. At JUNO’s construction site, the Earth Magnetic Field (EMF) is approximately 0.45 G [2]. Therefore, the PMTs of JUNO detector are necessary to be shielded from the EMF.This work aims to design current-carrying coils that generate magnetic field in the opposite direction to the EMF, thus, the two fields tend to cancel each other. We found that the magnetic field generated by the 32 circular coils forming a sphere of diameter 43.5 m can reduce the residual magnetic field to be lower than 10% of the EMF at the CD region and less than 20% at the veto region. Moreover, considering the secular change of the EMF’s inclination angle of approximately 2.87° over the 20 years of its operation, the coils can maintain the residual magnetic field both in the CD and the veto regions to be less than 10% and 20%, respectively.

Study on the radon removal for the water system of Jiangmen Underground Neutrino Observatory

The Jiangmen Underground Neutrino Observatory (JUNO), a 20-kton multipurpose underground liquid scintillator detector, was proposed with the determination of the neutrino mass hierarchy as a primary physics goal. To veto the cosmogenic background and shield the central detector from the environmental radioactivity, a multi-veto system, which consists of a water Cherenkov detector and a top tracker detector, is required. In order to keep the water quality good and remove the radon from water, an ultrapure water system, a radon removal system and radon concentration measurement system have been designed.

The design of the JUNO veto system

The Jiangmen Underground Neutrino Observatory (JUNO) is a multipurpose 20 kton liquid scintillator detector. The detector will be built in a 700 m deep underground laboratory, and its primary physics goal will be to determine the neutrino mass hierarchy. Due to the low background requirement of the experiment, a multi-veto system for cosmic muon detection and background reduction is designed. The volume outside the central detector is filled with pure water and equipped with 2000 MCP-PMTs (20 inches) to form a water Cherenkov detector for muon tagging. A Top Tracker system will be built by re-using the Target Tracker plastic scintillator modules of the OPERA experiment and will cover half of the top area. This will provide valuable information for cosmic muon induced 9Li/8He study.

Status of the JUNO reactor anti-neutrino experiment

The Jiangmen Underground Neutrino Observatory (JUNO) is a reactor antineutrino experiment with the aim to determine the neutrino mass hierarchy. The detector will be filled with 20 kilotons of liquid scintillator and instrumented with 18000 20-inch PMTs to achieve an unprecedented energy resolution of 3%@1 MeV. A 35.4 m diameter acrylic sphere will be built as a liquid scintillator vessel.The detector will be constructed in a 700-m-deep-underground laboratory to reduce cosmogenic muon flux. An external veto cosisting of a water Cherenkov detector and a top tracker will be used for cosmogenic muon detection and background reduction. The mass hierarchy sensitivity is expected to reach 3–4σ after 6 years of data taking. Civil construction and detector R&D are underway. Data taking is expected to start in 2020.

Citizen-Based Litter and Marine Debris Data Collection and Mapping

Marine debris and plastic in our oceans is a global issue of increasing concern. However, the monitoring of litter and debris is challenging at the global scale because of disconnected local organizations and the use of paper and pen for documentation. The Marine Debris Tracker mobile app and citizen science program allow for the collection of global standardized data at a scale, speed, and efficiency that wasn't previously possible. The app also serves as an outreach and education tool, with a Web portal that instantly shows the data that users have logged. User engagement through a top tracker competition and social media keeps people interested in the Marine Debris Tracker community. More than 400,000 items have been tracked, with plastic being the most common category noted; maps provide both global and local distribution of data. The Marine Debris Tracker community and dataset continue to grow daily.

The properties of glass resistive plate chambers made of different glasses

Glass resistive plate chambers (GRPCs) have been proposed as the basic element for the JUNO top tracker detector. With good uniform performance and low cost, GRPCs are well suited for large area experiments. Glass RPCs used in underground experiments require specially designed cassette and gas flow systems, since the glass is fragile and easily corroded by acid generated by water entering the gas-filled chamber. High-strength and chemical-resistant glasses have been proposed for underground experiments. We present here the test results of four GRPC chambers made of different glasses: normal thin glass, two high-strength glasses, and a chemical-resistant glass. The chemical-resistant and high-strength glasses have good surface quality, but their volume resistivities are higher. Higher resistivities lead to a higher required voltage to reach plateau operation, meaning that these glasses can only work in a very low rate experiment.