Local Ambiguities Research Articles

Enhanced prosody adds to morpho-syntactic cues in the interpretation of structural ambiguities in German

ABSTRACT This study investigated the effects of syntactically marked and enhanced prosody on local ambiguity resolution in German SVO and OVS sentences. In a visual-world experiment, thirty younger and thirty elderly healthy participants performed a sentence-picture matching task. Response accuracy, reaction times and fixations proportions to the target picture were analysed using linear mixed models. We found no support for beneficial effects of syntactically marked prosody, however, results suggested a facilitative role of enhanced prosodic cues (i.e. increased f0 maximum) prior to the point of disambiguation in SVO structures, as well as the beneficial effects of enhanced prosody adding to morpho-syntactic cues in OVS structures. Both age groups showed comparable cue use but inter-individual variability in prosodic cue processing. Overall, our study replicates and extends previous findings demonstrating the importance of examining variability in prosodic cue processing in future research.

InChI isotopologue and isotopomer specifications

This work presents a proposed extension to the International Union of Pure and Applied Chemistry (IUPAC) International Chemical Identifier (InChI) standard that allows the representation of isotopically-resolved chemical entities at varying levels of ambiguity in isotope location. This extension includes an improved interpretation of the current isotopic layer within the InChI standard and a new isotopologue layer specification for representing chemical intensities with ambiguous isotope localization. Both improvements support the unique isotopically-resolved chemical identification of features detected and measured in analytical instrumentation, specifically nuclear magnetic resonance and mass spectrometry.Scientific contributionThis new extension to the InChI standard would enable improved annotation of analytical datasets characterizing chemical entities, supporting the FAIR (Findable, Accessible, Interoperable, and Reusable) guiding principles of data stewardship for chemical datasets, ultimately promoting Open Science in chemistry.

Convolutional Cross-View Pose Estimation.

We propose a novel end-to-end method for cross-view pose estimation. Given a ground-level query image and an aerial image that covers the query's local neighborhood, the 3 Degrees-of-Freedom camera pose of the query is estimated by matching its image descriptor to descriptors of local regions within the aerial image. The orientation-aware descriptors are obtained by using a translationally equivariant convolutional ground image encoder and contrastive learning. The Localization Decoder produces a dense probability distribution in a coarse-to-fine manner with a novel Localization Matching Upsampling module. A smaller Orientation Decoder produces a vector field to condition the orientation estimate on the localization. Our method is validated on the VIGOR and KITTI datasets, where it surpasses the state-of-the-art baseline by 72% and 36% in median localization error for comparable orientation estimation accuracy. The predicted probability distribution can represent localization ambiguity, and enables rejecting possible erroneous predictions. Without re-training, the model can infer on ground images with different field of views and utilize orientation priors if available. On the Oxford RobotCar dataset, our method can reliably estimate the ego-vehicle's pose over time, achieving a median localization error under 1 m and a median orientation error of around 1 ° at 14 FPS.

Vision-inertial collaborative localization of multi-agents with remote interaction

In collaborative missions, it is crucial for each agent to have a clear understanding of the status among others. Therefore, various related works have been conducted, including virtual reality interactions and real-life swarm jobs. However, these types of collaborations still face certain limitations: (i) Limited distance: In virtual reality, users interact with others over a network within a predefined area and are unable to move over longer distances. (ii) Weak/non-GPS signals: Swarm jobs, which rely solely on localization information like GPS, often suffer from drift or interruptions. Moreover, significant location devices are rarely deployed in consumer-grade products. To address these challenges, we propose a multi-agent collaborative localization framework based on Camera, IMU (Inertial Measurement Unit), and a remote interaction module. This framework not only achieves globally consistent localization but also overcomes the distance limitations of WIFI or Bluetooth. Additionally, we tackle the issue of scale ambiguity in visual-inertial collaborative localization. Simulations and physical experiments validate the effectiveness of our proposed multi-agent collaboration with remote interaction. Unlike traditional applications that depend on external devices like Motion Capture Devices or GPS, our system achieves globally consistent localization internally.

ReeBundle: A Method for Topological Modeling of White Matter Pathways Using Diffusion MRI.

Tractography can generate millions of complex curvilinear fibers (streamlines) in 3D that exhibit the geometry of white matter pathways in the brain. Common approaches to analyzing white matter connectivity are based on adjacency matrices that quantify connection strength but do not account for any topological information. A critical element in neurological and developmental disorders is the topological deterioration and irregularities in streamlines. In this paper, we propose a novel Reeb graph-based method "ReeBundle" that efficiently encodes the topology and geometry of white matter fibers. Given the trajectories of neuronal fiber pathways (neuroanatomical bundle), we re-bundle the streamlines by modeling their spatial evolution to capture geometrically significant events (akin to a fingerprint). ReeBundle parameters control the granularity of the model and handle the presence of improbable streamlines commonly produced by tractography. Further, we propose a new Reeb graph-based distance metric that quantifies topological differences for automated quality control and bundle comparison. We show the practical usage of our method using two datasets: (1) For International Society for Magnetic Resonance in Medicine (ISMRM) dataset, ReeBundle handles the morphology of the white matter tract configurations due to branching and local ambiguities in complicated bundle tracts like anterior and posterior commissures; (2) For the longitudinal repeated measures in the Cognitive Resilience and Sleep History (CRASH) dataset, repeated scans of a given subject acquired weeks apart lead to provably similar Reeb graphs that differ significantly from other subjects, thus highlighting ReeBundle's potential for clinical fingerprinting of brain regions.

Deep Learning-Based Geomagnetic Navigation Method Integrated with Dead Reckoning

Accurate location information has significant commercial and economic value as they are widely used in intelligent manufacturing, material localization and smart homes. Magnetic sequence-based approaches show great promise mainly due to their pervasiveness and stability. However, existing geomagnetic indoor localization methods are facing the problems of location ambiguity and feature extraction deficiency, which will lead to large localization errors. To address these issues, we propose a coarse-to-fine geomagnetic indoor localization method based on deep learning. First, a multidimensional geomagnetic feature extraction method is presented which can extract magnetic features from spatial and temporal aspects. Then, a hierarchical deep neural network model is devised to extract more accurate geomagnetic information and corresponding location clues for more accurate localization. Finally, localization is achieved through a particle filter combined with IMU localization. To evaluate the performance of the proposed methods, we carried out several experiments at three trial paths with two heterogeneous devices, Vivo X30 and Huawei Mate30. Experimental results demonstrate that the proposed algorithm can achieve more accurate localization performance than the state-of-the-art methods. Meanwhile, the proposed algorithm has low cost and good pervasiveness for different devices.

An Analysis on Chinese EFL Learners’ Understanding of “Direct Object/Sentence Complement” Garden Path Sentences

This study explores the comprehension of Direct Object/Sentence Complement Garden Path sentences among Chinese EFL learners of different proficiency levels, as examined through online questionnaires featuring multiple introspective tests and judgment tasks. A total of 80 respondents participated in the survey, including 38 high proficiency level and 42 low proficiency level English EFL learners. The results indicated that due to a tendency among Chinese EFL learners to select interpretations with fewer attachment nodes, both groups exhibited garden path phenomena in sentence comprehension. Furthermore, following the garden path phenomena, both groups successfully resolved local ambiguities. However, the disambiguation accuracy was higher in the advanced group, corroborating the influence of language proficiency on the capability to reinterpret sentences. Ultimately, this research aims to further probe the underlying mechanisms in the comprehension of unique English sentence structures among learners, thus providing insights conducive to future English instruction and language learning.

On the Importance of Domain Awareness in Classifier Interpretations in Medical Imaging

Translating the success of deep learning-based computer-assisted classification into clinical adaptation hinges on the ability to explain a prediction's causality. Post-hoc interpretability approaches, especially counterfactual techniques, have shown both technical and psychological potential. Nevertheless, currently dominant approaches utilize heuristic, unvalidated methodology. Thereby, they potentially operate the underlying networks outside their validated domain, adding doubt in the predictor's abilities instead of generating knowledge and trust. In this work, we investigate this out-of-distribution problem for medical image pathology classifiers and propose marginalization techniques and evaluation procedures to overcome it. Furthermore, we propose a complete domain-aware pipeline for radiology environments. Its validity is demonstrated on a synthetic and two publicly available image datasets. Specifically, we evaluate using the CBIS-DDSM/DDSM mammography collection and the Chest X-ray14 radiographs. Our solution shows, both quantitatively and qualitatively, a significant reduction of localization ambiguity and clearer conveying results.

Localization Distillation for Object Detection.

Previous knowledge distillation (KD) methods for object detection mostly focus on feature imitation instead of mimicking the prediction logits due to its inefficiency in distilling the localization information. In this paper, we investigate whether logit mimicking always lags behind feature imitation. Towards this goal, we first present a novel localization distillation (LD) method which can efficiently transfer the localization knowledge from the teacher to the student. Second, we introduce the concept of valuable localization region that can aid to selectively distill the classification and localization knowledge for a certain region. Combining these two new components, for the first time, we show that logit mimicking can outperform feature imitation and the absence of localization distillation is a critical reason for why logit mimicking under-performs for years. The thorough studies exhibit the great potential of logit mimicking that can significantly alleviate the localization ambiguity, learn robust feature representation, and ease the training difficulty in the early stage. We also provide the theoretical connection between the proposed LD and the classification KD, that they share the equivalent optimization effect. Our distillation scheme is simple as well as effective and can be easily applied to both dense horizontal object detectors and rotated object detectors. Extensive experiments on the MS COCO, PASCAL VOC, and DOTA benchmarks demonstrate that our method can achieve considerable AP improvement without any sacrifice on the inference speed. Our source code and pretrained models are publicly available at https://github.com/HikariTJU/LD.

Progressive Perception Learning for Main Coronary Segmentation in X-ray Angiography.

Main coronary segmentation from the X-ray angiography images is important for the computer-aided diagnosis and treatment of coronary disease. However, it confronts the challenge at three different image granularities (the semantic, surrounding, and local levels). The challenge includes the semantic confusion between the main and collateral vessels, low contrast between the foreground vessel and background surroundings, and local ambiguity near the vessel boundaries. The traditional hand-crafted feature-based methods may be insufficient because they may lack the semantic relationship information and may not distinguish the main and collateral vessels. The existing deep learning-based methods seem to have issues due to the deficiency in the long-distance semantic relationship capture, the foreground and background interference adaptability, and the boundary detail information preservation. To solve the main coronary segmentation challenge, we propose the progressive perception learning (PPL) framework to inspect these three different image granularities. Specifically, the PPL contains the context, interference, and boundary perception modules. The context perception is designed to focus on the main coronary vessel based on the semantic dependence capture among different coronary segments. The interference perception is designed to purify the feature maps based on the foreground vessel enhancement and background artifact suppression. The boundary perception is designed to highlight the boundary details based on boundary feature extraction through the intersection between the foreground and background predictions. Extensive experiments on 1085 subjects show that the PPL is effective (e.g., the overall Dice is greater than 95%), and superior to thirteen state-of-the-art coronary segmentation methods.

Reducing ambiguities in single-hydrophone matched-field processing by exploiting source motion

Matched-field processing for localizing an underwater acoustic source in range and depth from power-spectrum measurements obtained at a single hydrophone receiver often suffers from high side-lobes and ambiguities when the signal is narrowband or signal bandwidth is small. In this paper we review how source motion can be utilized to reduce side-lobes in depth and range via an incoherent synthetic-aperture-like approach and explain the mechanisms responsible for side-lobe reduction relative to the height of the main-lobe by using a normal-mode expansion for the pressure field. We also derive an approximation for the depth main-lobe width when the true source range is known for the ideal rigid-waveguide case. Numerical results are presented corroborating the analytical analysis along with some matched-field localization ambiguity surface examples for (a) an ideal shallow-water waveguide with a pressure-release top boundary and a rigid bottom boundary and (b) a more realistic shallow-water Pekeris environment, to demonstrate how side-lobes and ambiguities are reduced when source motion is exploited in the matched-field processing.

Integrating AOA With TDOA for Joint Source and Sensor Localization

This paper investigates the use of angles of arrival (AOAs) in addition to time differences of arrival (TDOAs) for the simultaneous localization of sources and sensors. Based on the notation that both kinds of measurements come from the same set of sources and sensors, new corollaries that govern the rank properties of the matrices formed by the noise free observations are derived. The rank properties are exploited for the noise reduction of TDOAs, the recovery of times of flight (TOFs) and the joint localization of sources and sensors. The utilization of both types of measurements provides more effective TDOA enhancement and TOF recovery. It also enables a closed-form localization solution that is more resilient to noise and simpler to evaluate, compared to using either one. Furthermore, the relative localization ambiguity reduces to translation only. Simulations validate the usefulness of the rank properties and demonstrate the performance of the algorithms.

Target Detection and Location by Fusing Delay-Doppler Maps

The use of delay-Doppler maps (DDMs) measured by Global Navigation Satellite System Reflectometry (GNSS-R) for target detection and location is a hot issue because of the potentially global coverage and short revisit periods of GNSS-R satellite missions. Existing researches have explored the detection of oil rigs, oil spills, sea ice, and other targets with spaceborne GNSS-R. However, these researches may not fully consider the advantages of multiple GNSS-R satellites, for example, Cyclone Global Navigation Satellite System (CYGNSS) which has eight satellites. Two problems of using multiple GNSS-R satellites for detection and location are how to estimate the sea clutter that is used to cancel the clutter in DDMs and how to remove location ambiguity caused by the transformation from the delay-Doppler (DD) domain to the spatial domain. In this article, an oil rig is taken as an example for detection and location with multiple GNSS-R satellites. To distinguish the oil rig from sea clutter, this article uses subspace projection to estimate the clutter and subtract the estimation from the DDMs. Moreover, after the clutter cancellation, this article projects the potential oil rig coordinates in the DD domain into a uniform coordinate system and removes the location ambiguity based on multisatellites fusion. The measured DDMs collected by CYGNSS are employed to conduct experiments to validate the feasibility of the proposed methods for detection and location with DDMs. By subtracting the estimated sea clutter from DDMs, the potential oil rig can be indicated in the DD domain. Two location experiments can calculate the unambiguous coordinate of the oil rig with a deviation less than 4 km.

Characterizing Microearthquakes Induced by Hydraulic Fracturing With Hybrid Borehole DAS and Three-Component Geophone Data

Fluids injected during hydraulic fracturing (fracking) in unconventional shale oil and gas reservoirs, geothermal system enhancement, wastewater disposal, and carbon capture and storage can induce microearthquakes. The spatiotemporal distribution of induced earthquakes is often used to trace the growth of fractures in target layers and guides production. We analyze microseismicity behavior induced by fracking in the Montney Formation, one of the largest unconventional oil and gas reservoirs in North America. An optical fiber deployed in a horizontal well provides extensive spatial sampling and data coverage for microseismic imaging. We design median and F-k filters to predict instrumental and random noise and further suppress them by adaptive noise subtraction. An elliptical vertical transverse isotropic (VTI) velocity model is derived from the Backus-averaged well-log sonic data and is modified to match the microseismic wavefronts by a grid search. We image 41 previously cataloged microearthquakes recorded by DAS using Geometric-mean Reverse Time Migration. We find that the fiber geometry’s lack of 2D/3D variations increases the non-uniqueness of the image point location, and the P-wave particle motions derived from 3C geophones data can effectively eliminate the location ambiguity. The spatiotemporal distribution of our updated locations agrees with the fracking schedule. Predicted P- and S-wave traveltimes from the updated locations also match with the observed waveforms. Analyzing data sensitivity to source locations confirms the potential limitations imposed on source imaging by the geometry of borehole observations and shows that relocation accuracy is directionally dependent. We also investigate the feasibility of estimating source focal mechanisms using realistic DAS and geophone observations. Our study provides guidance for characterizing microearthquake sources and optimizing observation geometry for unconventional reservoirs.

Cross-scale rockmass characterization by high resolution seismic imaging around tunnels

Reflection seismic surveys are relevant to the detailed characterisation of crystalline bedrock. Relatively small-scale features, even single fractures, can demonstrably be detected. Integrated analysis and modelling of geological, hydrogeological and geophysical data is instrumental in the interpretation of reflection seismic data. This requires the inference of the nature of the reflectors, e.g. lithological contacts or deformation zones and resolving the orientation ambiguity of the reflectors. Brittle deformation zones and large fractures may be heterogeneous in their detailed structure, making their characterization a challenging task. Several examples of structure characterization ahead and around tunnels are presented, with focus on the safety assessment for engineering and mining applications. Seismic imaging techniques adapted for tunnels and galleries are presented with examples from high resolution seismic surveys carried out at ONKALO (Finland) and Äspö HRL (Sweden). Studies performed over a span of several years demonstrate that large fractures identified initially through geological mapping, can also be found, positioned and characterized by seismics at the small scale of tunnels and/or boreholes. We show that reflection seismic surveys are relevant to the detailed characterization of crystalline bedrock. The detection of some distinctive features, even large-scale, can be uncertain if the survey layout is spatially constrained. Combinations of borehole and tunnel measurements using measuring arrays with diverse orientations helped reduce the location ambiguities and are recommended to be used in mining and tunneling rock characterization surveys wherever possible.

Resolution of matched field processing for a single hydrophone in a rigid waveguide.

This paper studies resolution of matched field processing for locating, in range and depth, a broadband underwater acoustic source from data measured at a single hydrophone receiver. For the case of an ideal rigid shallow-water waveguide with a pressure-release top boundary and a rigid bottom boundary, the paper derives approximations for the main-lobe widths of the ambiguity surface. The two cases studied in this paper are (1) when coherent measurements of the pressure are available, with the transmitted source waveform precisely known, and (2) when only measurements of the received-signal pressure magnitude-squared are available, such as might occur when the transmitted signal is random and unknown. The analysis uses the normal-mode expansion for the pressure field to derive approximate expressions for the ambiguity-surface main-lobe widths, as a function of the number of modes and frequency band, for both range and depth. Numerical results are presented corroborating the analytical analysis. Finally, the paper argues that this ambiguity analysis can also give insights into Pekeris waveguides under appropriate conditions and shows numerical simulations of matched-field localization ambiguity surfaces for some realistic shallow-water Pekeris environments.

Shape-Former: Bridging CNN and Transformer via ShapeConv for multimodal image matching

As with any data fusion task, the front-end of the pipeline for image fusion, aiming to collect multitudinous physical properties from multimodal images taken by different types of sensors, requires registering the overlapped content of two images via image matching. In other words, the accuracy of image matching will influence directly the subsequent fusion results. In this work, we propose a hybrid correspondence learning architecture, termed as Shape-Former, which is capable of solving matching problems such as multimodal, and multiview cases. Existing attempts have trouble capturing intricate feature interactions for seeking good correspondence, if the image pairs simultaneously suffer from geometric and radiation distortion. To address this, our key is to take advantage of convolutional neural network (CNN) and Transformer for enhancing structure consensus representation ability. Specifically, we introduce a novel ShapeConv so that CNN and Transformer can be generalized to sparse matches learning. Furthermore, we provide a robust soft estimation of outliers mechanism for filtering the response of outliers before capturing shape features. Finally, we also propose coupling multiple consensus representations to further solve the context conflict problems such as local ambiguity. Experiments with variety of datasets reveal that our Shape-Former outperforms state-of-the-art on multimodal image matching, and shows promising generalization ability to different types of image deformations.

Patch-Based Uncalibrated Photometric Stereo Under Natural Illumination.

This paper presents a photometric stereo method that works with unknown natural illumination without any calibration objects or initial guess of the target shape. To solve this challenging problem, we propose the use of an equivalent directional lighting model for small surface patches consisting of slowly varying normals, and solve each patch up to an arbitrary orthogonal ambiguity. We further build the patch connections by extracting consistent surface normal pairs via spatial overlaps among patches and intensity profiles. Guided by these connections, the local ambiguities are unified to a global orthogonal one through Markov Random Field optimization and rotation averaging. After applying the integrability constraint, our solution contains only a binary ambiguity, which could be easily removed. Experiments using both synthetic and real-world datasets show our method provides even comparable results to calibrated methods.

Robust Vehicle Positioning Based on Multi-Epoch and Multi-Antenna TOAs in Harsh Environments

For radio-based time-of-arrival (TOA) positioning systems applied in harsh environments, obstacles in the surroundings and on the vehicle itself will block the signals from the anchors, reduce the number of available TOA measurements and thus degrade the localization performance. Conventional multi-antenna positioning technique requires a good initialization to avoid local minima, and suffers from location ambiguity due to insufficient number of TOA measurements and/or poor geometry of anchors at a single epoch. In this paper, taking advantage of the multi-epoch and multi-antenna (MEMA) TOA measurements bridged by inter-epoch constraints to utilize more information and improve the geometry of visible anchors, we propose a new positioning method, namely MEMA-TOA method. A new initialization method based on semidefinite programming (SDP), namely MEMA-SDP, is first designed to address the initialization problem of the MEMA-TOA method. Then, an iterative refinement step is developed to obtain the optimal positioning result based on the MEMA-SDP initialization. We derive the Cramér-Rao lower bound (CRLB) to analyze the accuracy of the new MEMA-TOA method theoretically, and show its superior positioning performance over the conventional single-epoch and multi-antenna (SEMA) localization method. Simulation results in harsh environments demonstrate that i) the new MEMA-SDP provides an initial estimation that is close to the real location, and empirically guarantees the global optimality of the final refined positioning solution, and ii) compared with the conventional SEMA method, the new MEMA-TOA method has higher positioning accuracy without location ambiguity, consistent with the theoretical analysis.

Compensating for Local Ambiguity With Encoder-Decoder in Urban Scene Segmentation

Semantic segmentation plays a critical role in scene understanding for self-driving vehicles. A line of efforts has proven that global context matters in urban scene segmentation due to massive scale changes. However, we find that existing methods suffer from local ambiguities when dissipating continuous local context, i.e. scrambling to a huge receptive field of global cues by coarse pooling. To this end, this paper proposes a new Context Aggregation Module (CAM) that consists of two primary components: context encoding using no coarse pooling but encoder-decoders with appropriate sampling scales and gated fusion that extends gate attention mechanism to balance different-scale context during feature fusion. Weeding out coarse pooling and applying the encoder-decoder inherits the merits of exploring global context while avoiding the drawback of losing local contextual continuity. We then construct a Context Aggregation Network (CANet) and conduct extensive evaluations on challenging autonomous driving benchmarks of Cityscapes, CamVid and BDD100K. Consistently improved results evidence the effectiveness. Notably, we attain competitive mIoU 82.7% on Cityscapes and optimal mIoU 80.5% on CamVid.