Segmentation Tracking Research Articles

Striatal dopamine explains novelty-induced behavioral dynamics and individual variability in threat prediction

Animals both explore and avoid novel objects in the environment, but the neural mechanisms that underlie these behaviors and their dynamics remain uncharacterized. Here, we used multi-point tracking (DeepLabCut) and behavioral segmentation (MoSeq) to characterize the behavior of mice freely interacting with a novel object. Novelty elicits a characteristic sequence of behavior, starting with investigatory approach and culminating in object engagement or avoidance. Dopamine in the tail of the striatum (TS) suppresses engagement, and dopamine responses were predictive of individual variability in behavior. Behavioral dynamics and individual variability are explained by a reinforcement-learning (RL) model of threat prediction in which behavior arises from a novelty-induced initial threat prediction (akin to "shaping bonus") and a threat prediction that is learned through dopamine-mediated threat prediction errors. These results uncover an algorithmic similarity between reward- and threat-related dopamine sub-systems.

Spatial representation by ramping activity of neurons in the retrohippocampal cortex

Neurons in the retrohippocampal cortices play crucial roles in spatial memory. Many retrohippocampal neurons have firing fields that are selectively active at specific locations, with memory for rewarded locations associated with reorganization of these firing fields. Whether this is the sole strategy for representing spatial memories is unclear. Here, we demonstrate that during a spatial memory task retrohippocampal neurons encode location through ramping activity that extends across segments of a linear track approaching and following a reward, with the rewarded location represented by offsets or switches in the slope of the ramping activity. Ramping representations could be maintained independently of trial outcome and cues marking the reward location, indicating that they result from recall of the track structure. When recorded in an open arena, neurons that generated ramping activity during the spatial memory task were more numerous than grid or border cells, with a majority showing spatial firing that did not meet criteria for classification as grid or border representations. Encoding of rewarded locations through offsets and switches in the slope of ramping activity also emerged in recurrent neural network models trained to solve a similar spatial memory task. Impaired performance of model networks following disruption of outputs from ramping neurons is consistent with this coding strategy supporting navigation to recalled locations of behavioral significance. Our results suggest that encoding of learned spaces by retrohippocampal networks employs both discrete firing fields and continuous ramping representations. We hypothesize that retrohippocampal ramping activity mediates readout of learned models for goal-directed navigation.

FarNet: An Attention-Aggregation Network for Long-Range Rail Track Point Cloud Segmentation

Rail track segmentation is key to environmental perception of autonomous train. However, due to the complexity of railway track environment, critical issues such as the detection of rail tracks with different curvatures remain to be overcome. In this study, a novel architecture called FarNet is proposed for long-range railway track point cloud segmentation. The proposed FarNet is mainly divided into three parts, i.e., spherical projection, attention-aggregation network and results refinement. Specifically, spherical projection converts the LiDAR point cloud into a pseudo range image, and attention-aggregation network enables railway track detection using the pseudo range image. Furthermore, in the attention-aggregation network two components, i.e., spatial attention module and information aggregation module, are proposed to enhance the capability of rail track segmentation. Last, the results refinement helps further filter out the noise points after segmentation. Experimental results show that the proposed FarNet achieved 98.0% mean intersection-over-union (MIoU) and 98.9% mean pixel accuracy (MPA) for rail track segmentation.

Data‐driven maintenance planning and scheduling based on predicted railway track condition

AbstractTimely planning and scheduling of railway infrastructure maintenance interventions are crucial for increased safety, improved availability, and reduced cost. We propose a data‐driven decision‐support framework integrating track condition predictions with tactical maintenance planning and operational scheduling. The framework acknowledges prediction uncertainties by using a Wiener process‐based prediction model at the tactical level. We also develop planning and scheduling algorithms at the operational level. One algorithm focuses on cost‐optimisation, and one algorithm considers the multi‐component characteristics of the railway track by grouping track segments near each other for one maintenance activity. The proposed framework's performance is evaluated using track geometry measurement data from a 34 km railway section in northern Sweden, focusing on the tamping maintenance action. We analyse maintenance costs and demonstrate potential efficiency increases by applying the decision‐support framework.

Piecewise model of track sinkage based on underwater cohesive soil

The soil resistance force prediction model plays a vital role in estimating trafficability of unmanned tracked vehicles on unknown terrain, especially for river floor soil, a kind of soft soil with high saturation. In order to obtain characteristics of soft soil, a soil test platform was built which track segment sinkage, and shear tests can be performed in river floor soil. Comprehensive analysis of mechanics of track-soil interaction, a new pressure-sinkage empirical model is proposed for soft soil. This piecewise empirical model with two functions relies on two sets of independent parameters to describe the stratification of the underwater soil, including loose layer and dense layer. To evaluate the accuracy of the piecewise empirical model, an underwater tracked vehicle driven by two brushless DC motors was designed and an underwater travel test platform was built, which the vehicle can be operating in the river floor soil. The main external forces enacting on the vehicle, including environmental loads caused by the river floor soil and water, were analyzed in the process of vehicle running at a constant speed, and a soil resistance force prediction model was deduced with the proposed piecewise empirical model. The current of two brushless DC motors is used to estimate the external loads and validate the piecewise empirical model. Test results indicate that the soil resistance force prediction model was effective and feasible.

A joint model and data driven track segment association algorithm for manoeuvring target tracking

In the field of manoeuvring target tracking, track breakage is a common issue due to strong manoeuvrability, missed detection, large measurement error, and long sampling interval, etc. In order to improve the track continuity of manoeuvring targets, a joint model and data-driven track segment association (JMDD-TSA) algorithm is proposed. First, the interacting multiple model (IMM) smoother is invoked to pre-process the track segments to obtain a smoothed state sequence. And the mode switching time (MST) is estimated by the IMM when a motion mode changing occurs in a track segment. Second, the estimated MST is used to select the partial smoothed states as the training data and the local Gaussian process (LGP) is proposed to predict and backtrack the track segments. Finally, a two-step association strategy is designed to assign and mend the candidate track segment pairs. Compared with the model-driven TSA and the data-driven TSA, the proposed algorithm can effectively improve the average track life and tracking accuracy of strong manoeuvring targets.

Track Segment Association Method Based on Bidirectional Track Prediction and Fuzzy Analysis

Due to sensor characteristics, geographical environment, electromagnetic interference, electromagnetic silence, information countermeasures, and other reasons, the phenomenon of track breakages occur in the process of aircraft track data processing. It leads to the change in target label attributes. In order to make the track segment association effect better, we studied several existing time series prediction methods, and proposed a track segment association method based on bidirectional Holt-Winters prediction and fuzzy analysis. This algorithm bidirectionally predicts and extrapolates track segments by the Holt-Winters method, and then uses the fuzzy track segment association algorithm to perform segment association and secondary association. The simulation results of this method show that the track segment association method based on Holt-Winters prediction and fuzzy analysis can effectively solve the track association problem where the target label attributes change before and after track breakage, demonstrating better association ability and robustness. Compared with the fuzzy association method without adding track prediction, our method generally improves the association accuracy by 35%.

Study of Clustering Algorithms in Object Tracking and Image Segmentation

Mean Shift is a kind of clustering algorithm, which is mostly used for target tracking, image segmentation, etc. In order to solve the problem that image information is not effectively utilized because of unclear traffic video images and random jitter between image sequences, this paper has studied how to achieve stability of traffic video images and proposed an improved Mean Shift algorithm about how to conduct object centroid registration in compensation for the deviations in space localization and, on this basis, how to select kernel window width to eliminate the errors in scale positioning. This algorithm gets the tracking effect and computation analysis of improved Mean Shift from the perspective of applications, removes or relieves the impact motion has on imaging, improves the quality of the video image information obtained, automatically adjusts the size of window according to the scale changes of moving object in the image, and effectively enhances the stability and real time of object tracking. Besides, in the postprocessing stage, the superpixel based on the Mean Shift algorithm is applied to further optimize the segmentation result; it is a popular mode-seeking clustering algorithm, which makes Mean Shift method ideal on low-dimensional applications such as image segmentation. Finally, we show promising results for remote sensing image segmentation.

Tracking Highly Similar Rat Instances under Heavy Occlusions: An Unsupervised Deep Generative Pipeline.

Identity tracking and instance segmentation are crucial in several areas of biological research. Behavior analysis of individuals in groups of similar animals is a task that emerges frequently in agriculture or pharmaceutical studies, among others. Automated annotation of many hours of surveillance videos can facilitate a large number of biological studies/experiments, which otherwise would not be feasible. Solutions based on machine learning generally perform well in tracking and instance segmentation; however, in the case of identical, unmarked instances (e.g., white rats or mice), even state-of-the-art approaches can frequently fail. We propose a pipeline of deep generative models for identity tracking and instance segmentation of highly similar instances, which, in contrast to most region-based approaches, exploits edge information and consequently helps to resolve ambiguity in heavily occluded cases. Our method is trained by synthetic data generation techniques, not requiring prior human annotation. We show that our approach greatly outperforms other state-of-the-art unsupervised methods in identity tracking and instance segmentation of unmarked rats in real-world laboratory video recordings.

DeepStrain Evidence of Asymptomatic Left Ventricular Diastolic and Systolic Dysfunction in Young Adults With Cardiac Risk Factors.

PurposeTo evaluate if a fully-automatic deep learning method for myocardial strain analysis based on magnetic resonance imaging (MRI) cine images can detect asymptomatic dysfunction in young adults with cardiac risk factors.MethodsAn automated workflow termed DeepStrain was implemented using two U-Net models for segmentation and motion tracking. DeepStrain was trained and tested using short-axis cine-MRI images from healthy subjects and patients with cardiac disease. Subsequently, subjects aged 18–45 years were prospectively recruited and classified among age- and gender-matched groups: risk factor group (RFG) 1 including overweight without hypertension or type 2 diabetes; RFG2 including hypertension without type 2 diabetes, regardless of overweight; RFG3 including type 2 diabetes, regardless of overweight or hypertension. Subjects underwent cardiac short-axis cine-MRI image acquisition. Differences in DeepStrain-based left ventricular global circumferential and radial strain and strain rate among groups were evaluated.ResultsThe cohort consisted of 119 participants: 30 controls, 39 in RFG1, 30 in RFG2, and 20 in RFG3. Despite comparable (>0.05) left-ventricular mass, volumes, and ejection fraction, all groups (RFG1, RFG2, RFG3) showed signs of asymptomatic left ventricular diastolic and systolic dysfunction, evidenced by lower circumferential early-diastolic strain rate (<0.05, <0.001, <0.01), and lower septal circumferential end-systolic strain (<0.001, <0.05, <0.001) compared with controls. Multivariate linear regression showed that body surface area correlated negatively with all strain measures (<0.01), and mean arterial pressure correlated negatively with early-diastolic strain rate (<0.01).ConclusionDeepStrain fully-automatically provided evidence of asymptomatic left ventricular diastolic and systolic dysfunction in asymptomatic young adults with overweight, hypertension, and type 2 diabetes risk factors.

Distinguishing Volcanic Contributions to the Overlapping Samoan and Cook-Austral Hotspot Tracks

Abstract To deconvolve contributions from the four overlapping hotspots that form the “hotspot highway” on the Pacific plate—Samoa, Rarotonga, Arago-Rurutu, and Macdonald—we geochemically characterize and/or date (by the 40Ar/39Ar method) a suite of lavas sampled from the eastern region of the Samoan hotspot and the region “downstream” of the Samoan hotspot track. We find that Papatua seamount, located ~60 km south of the axis of the Samoan hotspot track, has lavas with both a HIMU (high μ = 238U/204Pb) composition (206Pb/204Pb = 20.0), previously linked to one of the Cook-Austral hotspots, and an enriched mantle I (EM1) composition, which we interpret to be rejuvenated and Samoan in origin. We show that these EM1 rejuvenated lavas at Papatua are geochemically similar to rejuvenated volcanism on Samoan volcanoes and suggest that flexural uplift, caused by tectonic forces associated with the nearby Tonga trench, triggered a new episode of melting of Samoan mantle material that had previously flattened and spread laterally along the base of the Pacific plate under Papatua, resulting in volcanism that capped the previous HIMU edifice. We argue that this process generated Samoan rejuvenated volcanism on the older Cook-Austral volcano of Papatua. We also study Waterwitch seamount, located ~820 km WNW of the Samoan hotspot, and provide an age (10.49 ± 0.09 Ma) that places it on the Samoan hotspot trend, showing that it is genetically Samoan and not related to the Cook-Austral hotspots as previously suggested. Consequently, with the possible exception of the HIMU stage of Papatua seamount, there are currently no known Arago-Rurutu plume-derived lava flows sampled along the swath of Pacific seafloor that stretches between Rose seamount (~25 Ma) and East Niulakita seamount (~45 Ma), located 1400 km to the west. The “missing” ~20-million-year segment of the Arago-Rurutu hotspot track may have been subducted into the northern Tonga trench, or perhaps was covered by subsequent volcanism from the overlapping Samoan hotspot, and has thus eluded sampling. Finally, we explore tectonic reactivation as a cause for anomalously young volcanism present within the western end of the Samoan hotspot track.

Particle detection and tracking with DNA

We present the first proof-of-concept simulations of detectors using biomaterials to detect particle interactions. The essential idea behind a “DNA detector” involves the attachment of a forest of precisely-sequenced single or double-stranded nucleic acids from a thin holding layer made of a high-density material. Incoming particles break a series of strands along a roughly co-linear chain of interaction sites and the severed segments then fall to a collection area. Since the sequences of base pairs in nucleic acid molecules can be precisely amplified and measured using polymerase chain reaction (PCR), the original spatial position of each broken strand inside the detector can be reconstructed with nm precision. Motivated by the potential use as a low-energy directional particle tracker, we perform the first Monte Carlo simulations of particle interactions inside a DNA detector. We compare the track topology as a function of incoming direction, energy, and particle type for a range of ionising particles. While particle identification and energy reconstruction might be challenging without a significant scale-up, the excellent potential angular and spatial resolution (lesssim 25^circ axial resolution for keV-scale particles and nm-scale track segments) are clear advantages of this concept. We conclude that a DNA detector could be a cost-effective, portable, and powerful new particle detection technology. We outline the outstanding experimental challenges, and suggest directions for future laboratory tests.

A Multi-Stage Vessel Tracklet Association Method for Compact High-Frequency Surface Wave Radar

A compact high-frequency surface wave radar, used for target detection, suffers from a low signal-to-noise ratio, low detection probability, a high false alarm rate, and low positioning accuracy; this is due to its low transmit power and the reduced aperture size of the receiving antenna array. When target tracking algorithms are applied to compact high-frequency surface wave radar data, track fragmentation often occurs and a long track may be broken into several track segments (a.k.a. tracklets), which degrade the tracking continuity for a maritime surveillance system. We present a multi-stage vessel tracklet association method, based on bidirectional prediction and optimal assignment, to associate the broken tracklets belonging to the same target, and connect them to form one continuous track in a multi-target tracking scenario. Firstly, two global motion parameters, i.e., the average heading and average speed, were, respectively, extracted from the newly initiated and terminated tracklets as features for a rough tracklet association, then k-means clustering was used to produce the preliminary tracklet pairs. Subsequently, the temporal and spatial constraints on the initiated and terminated tracklets were considered to refine the preliminary tracklet pairs, to obtain the candidate tracklet pairs. Finally, the tracklet association costs were calculated using Doppler velocity, range, and azimuth to determine the similarity between tracklets in the candidate tracklet pairs, and an association cost matrix was obtained. Then an optimal assignment method based on Jonker–Volgenant–Castanon algorithm was applied to the association matrix to achieve optimal tracklet matching by minimizing the total association costs. Tracklet association experiments with both simulated and field data were conducted; experimental results show that, compared with existing track segment association methods, the association accuracy of the proposed method is significantly improved with better tracking continuity.

Graph Neural Networks for Charged Particle Tracking on FPGAs.

The determination of charged particle trajectories in collisions at the CERN Large Hadron Collider (LHC) is an important but challenging problem, especially in the high interaction density conditions expected during the future high-luminosity phase of the LHC (HL-LHC). Graph neural networks (GNNs) are a type of geometric deep learning algorithm that has successfully been applied to this task by embedding tracker data as a graph—nodes represent hits, while edges represent possible track segments—and classifying the edges as true or fake track segments. However, their study in hardware- or software-based trigger applications has been limited due to their large computational cost. In this paper, we introduce an automated translation workflow, integrated into a broader tool called hls4ml, for converting GNNs into firmware for field-programmable gate arrays (FPGAs). We use this translation tool to implement GNNs for charged particle tracking, trained using the TrackML challenge dataset, on FPGAs with designs targeting different graph sizes, task complexites, and latency/throughput requirements. This work could enable the inclusion of charged particle tracking GNNs at the trigger level for HL-LHC experiments.

Simulation of Track-Soft Soil Interactions Using a Discrete Element Method

With the development of unmanned tracked vehicles, soil model predictions of soft terrains are becoming more essential. In order to accurately simulate the interaction characteristics between soil particles and the track, soil modeling with a discrete element method (DEM) is proposed. Volume-based scaled-up modeling of DEM soil particles and the calibration of DEM input parameters were investigated as a feasible approach to realizing many particle calculations. Calibration of DEM input parameters can solve the distortions between actual and DEM particle sizes. Cohesion and friction parameters of the scaled-up soil particle model were recalibrated by the shape accumulated through the virtual design of the experiment. Soil DEM particles were scaled up to 1 mm spherical particles, and recalibrated DEM parameter values were used to match the actual accumulated soil shape. Three calibrated scaled-up soil models were used for the shear stress–displacement DEM simulation of a track segment, and the mean absolute percentage error (MAPE) was less than 11% compared with the actual shear stress–displacement test. The parameter value of soil traction performance empirical model of a tracked vehicle is modified according to the soil shear stress–displacement DEM simulation. Comparative analysis was performed for travel test results of a tracked vehicle; the relative error of the soil traction prediction results to actuals was less than 16.8%. This showed that the volume-based particle scaling technique is an effective DEM for the mechanical simulation of soil.

Track Segment Association With Dual Contrast Neural Network

Track interruption will cause serious interference to target tracking, situation awareness, and information fusion. And it is often triggered by a variety of reasons, such as target maneuver, the blind spot of Doppler radar, and excessive long sampling interval, among other causes. All the existing algorithms are based on the hypothetical target motion models and adopt the theory of statistical estimation to measure the similarity between predicted tracks to associate, which have some disadvantages such as unreasonable hypotheses, inapplicable models, and uncertain thresholds. In order to address these problems, in this article, we propose a track segment association dual contrast neural network (TSADCNN) to complete track segment association tasks. First, based on deep metric learning, we propose a temporal and spatial information extraction module to extract track information, especially symmetric correlation information and raise dimension. Second, by constructing an appropriate optimization function, TSADCNN makes track segments belonging to the same target closer, track segments belonging to different targets farther in high-dimensional space. Finally, TSADCNN will pick the nearest neighbor vectors in high-dimensional space as associated tracks. Extensive experiments on simulation data are conducted to evaluate the effectiveness of TSADCNN, and the results demonstrate that this method has good adaptability and anti-noise ability. Furthermore, the contrast experiment shows that it can outperform other TSA methods in both quality and efficiency. Compared with the best performance, the average true association rate of TSADCNN can be increased by 1.3% in long interrupt intervals.

LTST: Long-term segmentation tracker with memory attention network

Recent interest in the combination of visual object tracking (VOT) and video object segmentation (VOS) has yielded rapid progress. However, existing segmentation methods are still restricted to the target model created in the first frame, which leads to the lack of long-term adaptability. To overcome this limitation, we propose a novel long-term segmentation tracker LTST leveraging memory attention network to achieve the effect of online learning without additional training. Specifically, we first combine a discriminative correlation filter with a matching-based paradigm for the segmentation task, then we develop a memory attention network based on partial cost volume to extract relevant historical information and dynamically reform the segmentation template. Moreover, we extend our LTST for long-term tracking by introducing a multi-scale verification network to identify tracking failures, and a global detector to re-locate the missing target. Experimental results on VOT-LT2018, VOT-LT2019, LaSOT and TLP benchmarks show that our proposed tracker achieves comparable performance to the state-of-the-art long-term tracking algorithms.

Semi-Supervised Deep Learning in High-Speed Railway Track Detection Based on Distributed Fiber Acoustic Sensing.

High deployment costs, safety risks, and time delays restrict traditional track detection methods in high-speed railways. Therefore, approaches based on optical sensors have become the most remarkable strategy in terms of deployment cost and real-time performance. Owing to the large amount of data obtained by sensors, it has been proven that deep learning, as a powerful data-driven approach, can perform effectively in the field of track detection. However, it is difficult and expensive to obtain labeled data from railways during operation. In this study, we used a segment of a high-speed railway track as the experimental object and deployed a distributed optical fiber acoustic system (DAS). We propose a track detection method that innovatively leverages semi-supervised deep learning based on image recognition, with a particular pre-processing for the dataset and a greedy algorithm for the selection of hyper-parameters. The superiority of the method was verified in both experiments and actual applications.

A Comprehensive Review for Typical Applications Based Upon Unmanned Aerial Vehicle Platform

Unmanned Aerial Vehicles (UAVs) have been widely applied in military and civilian fields due to their flexibility and effectiveness. As a vital component of UAVs, the vision system has taken on great significance in different applications (e.g., autonomous landing, traffic surveillance, and disaster rescue) to attract widespread attention in recent years. Therefore, the automatic understanding of visual data collected from these air platforms becomes urgently needed in UAV systems. In this review, we revisit and summarize the recent techniques and developments for several typical UAV applications, including object detection, object tracking, and semantic segmentation. In addition, we also highlight the difficulties and subsequent orientations from different perspectives, which may stimulate future research and applications in the UAV vision era.

Adaptive Multipoint Method for Predicting Geometric Parameters of the Railway Track Based on Convergence Theory

The article describes the properties and application of the original adaptive multipoint (for each irregularity size) method for predicting the values of track geometric parameters and their changes over time, based on the use of convergence theory. Prediction results are presented in the form of irregularity size distribution function (ISDF). ISDF shows the cumulative length of specific-size track irregularity within an arbitrary length track segment at the end of the prediction interval. The method is based on three main principles: using the ISDF function to describe track condition and changes therein, using only the results of previous measurements for calculations as input information about the condition of the track, using the ISDF convergence process analysis to calculate future values of the track geometric parameters. The method is invariant to the length of the time interval between past measurements. The method also allows to identify a tendency to sudden spontaneous deterioration of the track, which does not follow from a regular trend. For longitudinal level defects, the average prediction error for defect sizes s>|3|mm and prediction intervals of 2 and 6 months does not exceed 0.35%.