IntroductionPre-anesthetic assessment involves a series of evaluations for airway management. However, the lack of quantitative analysis may limit its reliability. In this study, we propose an optimized YOLO11-Pose framework with a DeBiFormer module for automated, image-based pre-anesthetic assessment.MethodsThe proposed model performs single-stage inference to localize clinically relevant facial and hand keypoints associated with mouth opening, thyromental distance, and neck mobility. The framework was evaluated on a dataset collected from Ruijin Hospital under controlled acquisition conditions.ResultsExperimental results demonstrated high detection performance. Measurement-based evaluations, including Bland-Altman analysis, pixel-level error, and normalized mean error (NME), indicated reliable agreement with reference annotations. In addition, Mallampati classification was evaluated to assess clinical applicability. The model achieved an accuracy of 77.34% in the four-class setting and improved performance in a clinically motivated binary setting, with an accuracy of 83.99% and a quadratic weighted kappa of 0.65, indicating substantial agreement.DiscussionThese results suggest that the proposed method provides robust and clinically meaningful assessment of airway characteristics and may support automated pre-anesthetic evaluation in clinical practice.

Accurate trajectory prediction of unmanned aerial vehicles (UAVs) is crucial for effective anti-UAV defense. However, existing methods are typically developed under ideal conditions and fail to maintain robustness under diverse disturbances. To address this challenge, we propose a Teacher-Student framework for UAV state estimation and trajectory forecasting that enhances reliability across diverse disturbances. The framework integrates diffusion-based denoising and audio-visual feature fusion to extract robust motion states, while pseudo-state supervision is derived from kinematic modeling and CAD-guided pose estimation. Experimental results demonstrate that our method consistently outperforms state-of-the-art baselines across both ideal and disturbed scenarios, achieving accurate long-horizon predictions essential for real-world anti-UAV applications. Code will be released to support future research in robust anti-UAV systems at https://github.com/hxlei0827/Robust-Anti-UAV-Under-Diverse-Disturbances.

Reliable 3D trajectory estimation of unmanned aerial vehicles (UAVs) is a fundamental requirement for anti-UAV systems, yet the acquisition of large-scale and accurately annotated trajectory data remains prohibitively expensive. In this work, we present a novel framework that derives UAV 3D trajectories and category information directly from Internet-scale UAV videos, without relying on manual annotations. First, language-driven data acquisition is employed to autonomously discover and collect UAV-related videos, while vision-language reasoning progressively filters task-relevant segments. Second, a training-free cross-modal label generation module is introduced to infer 3D trajectory hypotheses and UAV type cues. Third, a physics-informed refinement process is designed to impose temporal smoothness and kinematic consistency on the estimated trajectories. The resulting video clips and trajectory annotations can be readily utilized for downstream anti-UAV tasks. To assess effectiveness and generalization, we conduct zero-shot transfer experiments on a public, well-annotated 3D UAV benchmark. Results reveal a clear data scaling behavior: as the amount of online video data increases, zero-shot transfer performance on the target dataset improves consistently, without any target-domain training. The proposed method closely approaches the current state-of-the-art, highlighting its robustness and applicability to real-world anti-UAV scenarios. Code and datasets will be released upon acceptance.

This paper focuses on the challenge of searching a shared feature space for face and voice modalities. Studies of human intelligence have shown that people can link faces and voices. However, computational intelligence research has paid less attention to the relationship between voice and face. The objective of this work is to identify generic features that are generalized to both visual and audio modalities, enabling the matching of faces and voices. To achieve a better approximation of the feature spaces of two modalities, a multi-level alignment approach is applied to their features. For individual sample pairs, a contrast learning approach is exploited. For the overall feature distribution, an optimal transport method is used. The impact of pre-trained weights on this task is explored. Compared to the simple contrastive learning method, inclusion of overall feature alignment improves both verification and matching accuracy. Extensive experimental results suggest that the overall distribution alignment is useful for the cross-model feature matching task. Also, the use of pre-trained parameters can improve the results in certain situations.

Recent years have witnessed the popularity of integrating Siamese network into RGBT tracking for fast-tracking. However, these trackers mostly utilize the feature information of the last output layer and ignore the benefits of multi-layer information. In addition, they often adopt feature-level fusion for different modalities but fail to explore the strength of decision-level fusion, which may easily decrease their flexibility and independence. In this paper, a novel multi-layer attention aggregation Siamese network on the decision level is proposed for robust RGBT tracking. To be specific, a hierarchical channel attention Siamese network is built to recalibrate the extracted multi-layer features from RGB and thermal infrared images. This can focus on more discriminative features to learn robust feature representation. Then, a depth-wise correlation operation is performed to produce RGB and thermal response maps, respectively. To better exploit and utilize the complementary RGB and thermal information, a contribution-aware aggregation network is designed to adaptively aggregate them. Lastly, a classification and regression network is adopted to complete the bounding box prediction. Extensive experiments on four large-scale RGBT benchmarks demonstrate outstanding tracking ability over other state-of-the-art trackers.

The integration of Artificial Intelligence (AI) and sign language recognition is a hot topic in the field of AI+Science, aimed at addressing communication barriers faced by the deaf and hard-of-hearing communities. This paper examines the integration of AI with sign language recognition, highlighting its potential to bridge communication gaps for the deaf and hard-of-hearing. It reviews the evolution of sign language recognition from data gloves to computer vision and underscores the role of extensive databases. The paper also discusses the benefits of multi-modal AI models in enhancing recognition accuracy. It highlights the importance of government and industry support, ethical data practices, and user-centered design in advancing this technology. The challenges and opportunities of integrating this technology into daily life, including technical, interface, and ethical considerations, are explored, emphasizing the need for user-focused solutions and innovative technical approaches.

The inferior sample efficiency of reinforcement learning (RL) and the requirement for high-quality demonstrations in imitation learning (IL) will hinder their application in real-world robots. To address this challenge, a novel self-evolution framework, named task-oriented self-imitation learning (TOSIL), is proposed. To circumvent external demonstrations, the top-K self-generated trajectories are chosen as expert data from both per-episode exploration and long-term return perspectives. Each transition is assigned a guide reward, which is formulated by these trajectories. The guide rewards update as the agent evolves, encouraging good exploration behaviors. This methodology guarantees that the agent explores in the direction relevant to the task, improving sample efficiency and asymptotic performance. The experimental results on locomotion and manipulation tasks indicate that the proposed framework outperforms other state-of-the-art RL methods. Furthermore, the integration of suboptimal trajectories has the potential to improve the sample efficiency while maintaining performance. This is a significant advancement in autonomous skill acquisition for robots.

Recently have testified the superior tracking ability of Transformer in RGBT tracking for its global and dynamic modeling property. However, these Transformer-based trackers lack attention to the primary feature information and are susceptible to interference from background information. In addition, they often either focus on shared modality information or specific modality information but fail to adequately explore the potential of these two patterns together. To address these issues, a sparse trifurcate Transformer aggregation network is proposed in this article for enhancing tracking robustness. First, a trifurcate tree structure is designed to obtain both modality-shared and modality-specific information, which can learn more powerful feature representations. Second, a sparse attention mechanism is adopted in Transformer to focus on the important features. To fully mine the complementary multimodal information, a confidence-aware aggregation network is designed to generate reliability weights of each mode. Finally, a double-head network is introduced to locate target. Sufficient experimental results on multiple RGBT benchmarks, including GTOT, RGBT210, RGBT234, and LasHeR, verify superior tracking ability against other advanced trackers.

In pattern recognition tasks, the information from system input is modeled through a series of nonlinear operations, which include but not limited to feature extraction, regression, and classification. Both theoretically and practically, these operations are inevitably subject to internal modeling error and external disturbance, resulting at a performance challenge. Those state-of-the-art methods, e.g. Convolutional Neural Network and Transformer, still display significant instabilities and failures under practical applications, so comes a lack of generalization. Consequently, the more robust pattern recognition methods and related theories still merit a further study. This paper firstly reviews those state-of-the-art technologies in the field. The bottleneck of performances in those latest researches is associated with a lack of disturbance estimation and corresponding compensation. Therefore, the implications of disturbance rejection in pattern recognition field are further discussed from a control point of view. Then, the open problems are summarized. Ultimately, a discussion of the potential solutions, which is related to the application of compensation on features, is given to highlight the future study. Through the systematic review in this paper, the disturbance rejection in pattern recognition is developed into a control problem. Hopefully, more effective control technologies for the compensation on features can be used to improve the robustness of pattern recognition theoretically and practically.

Recent years have witnessed the exciting performance of trackers based on Transformer. However, they usually separate the process of information extraction and integration, weakening the information interaction between the target and search region. In addition, they depend on traditional Transformer to model the long range dependency, which leads to a lack of focus on the primary information needed by high-accuracy trackers. In this paper, a sparse mixed attention aggregation model is proposed for robust tracking based on visible and thermal infrared images. To be specific, a backbone network composed of sparse mixed attention is designed to achieve information extraction and integration. This is helpful to obtain specific discriminative feature information and enhance their communication. To give full play to the complementary visible and thermal information, a confidence aware aggregation network is designed, which can learn the reliable confidence of visible and thermal branches. Finally, a corner-based localization head is introduced to estimate the target state. Extensive experiments on three large-scale multimodal tracking benchmarks demonstrate the superior tracking ability of the proposed tracker over other advanced trackers.

Recent years have witnessed neuromorphic vision sensor (NVS) driving the performance of dense prediction in the domain of visual perception because of its unique properties. Although prior works have designed elaborate pipelines to solve different dense prediction problems, they do not consider the large amount of information synergies among tasks, which are compensated by complex models as well as large amounts of labeled data. To this end, we propose to exploit inter-task synergies to improve the performance of dense predictions in the NVS domain and introduce the first multitask learning (MTL) model for NVS. Specifically, our NVS-oriented MTL model employs a hard parameter sharing scheme. A shared encoder is adopted to extract a universal representation from the input event streams, which is then branched into several task-specific decoders for separating domain-specific information within individual tasks. To account for the distinctive nature of event streams, a hierarchical recurrent vision Transformer (RViT) backbone is proposed as the shared encoder part. It is capable of modeling both global and local spatial context from sparse event signals while also leveraging the temporal cues within. Extensive evaluations are conducted on a recent NVS benchmark (DSEC) to verify our model. Our method outperforms all baselines as well as the state-of-the-art (SOTA) MTL networks dedicated to conventional cameras on two dense prediction tasks simultaneously. In addition, extensive ablations demonstrate the effectiveness of our architectural design and component selections. We believe our research paves the way for the introduction of the MTL strategy in the field of event-based vision.

In recent years, visual object tracking, as a prominent research area in computer vision, has garnered significant attention. To bolster the robustness of trackers across a spectrum of complex scenarios, researchers actively explore the synergistic potential of visible and thermal infrared images, aiming to design more potent tracking systems. This paper presents a comprehensive review of target tracking technology based on visible and thermal infrared information, encompassing three key aspects. Firstly, we categorize existing RGBT tracking methods into two main categories: traditional-based methods and deep learning-based methods. This classification facilitates a systematic understanding and comparison of the strengths and weaknesses of different approaches, providing a solid foundation for future research. Secondly, we focus on the evolution of RGBT datasets and analyze the performance of diverse tracking methods on these datasets. Research in this domain aids in evaluating the applicability of existing methods in real-world scenarios and offers guidance for future dataset construction. Finally, we delve into future research directions from multiple perspectives, including model design and dataset construction. In terms of model design, researchers are encouraged to explore more efficient feature extraction methods and innovative model fusion structures to further enhance tracker performance. Regarding dataset construction, increased attention should be given to ensure diversity in real-world scenarios, guaranteeing optimal tracker performance across a variety of complex conditions. In conclusion, this review makes a comprehensive analysis of the development of RGBT tracking from different perspectives, and provides a valuable reference for researchers in related fields such as multi-modal tracking and image fusion. By systematically classifying and analyzing existing research while outlining future research prospects, this review aims to foster the continued development of this field and inspire the emergence of more innovative work.

In this paper, an active disturbance rejection predictive control strategy is proposed for trajectory planning task of unmanned ground vehicles. Rather than error estimation of accurate system modeling, internal error and environment disturbance are processed via single extended state observer. Nonlinear feedback control law is applied to reduce steady-state error significantly. Then motion planning and nonholonomic constraints could be handled via nonlinear model predictive control. Simulation and experiment results show that the proposed algorithm is able to accomplish trajectory planning task with internal error and environment disturbance.

The complete feature set is one of the challenging topics in the pattern recognition field. Lots of researches are taken on how to obtain the robust features which can be transferred into different kinds of applications, i.e. an excellent generality. However, those features in the state-of-the-art recognition model are usually redundant. The internal modeling errors in network architecture, and the contaminated training samples, leave the features beyond Complete. Aiming at the completeness problem, this paper presents a complete feature point of view. A group of decoupled features are obtained by the adversarial learning framework. Then, these features are used to construct the complete feature set, which not only can be used in a limited source scenario, but also perform well in cross dataset verification. Then, the proposed method is examined in applications of knowledge distillation and cross domain facial recognition task. The experimental results demonstrate an effectiveness of the proposed method. The construction of complete feature set can bring a more robust pattern recognition model.

In the field of artificial intelligence, pattern recognition is widely used to extract the abstract information in those high dimensional inputs of image, voice, or video. However, the interpretability of pattern recognition still remains understudied. The incomplete features extracted from system input still limit the recognition performance. To reject the disturbance of feature incompleteness, an error compensation is realized into the pattern recognition model under a quantum computation framework. The quantum-based recognition system fulfills the information transmission from input to output with the transformation of quantum states. Then, a compensation for the quantum state is used to reject those intermediate errors in the pattern recognition task. The experiment results in this paper indicate an effectiveness of the proposed method, with which the compensated Quantum Neural Network obtains a better performance. The proposed method brings a more robust recognition system under unknown disturbances.

This paper integrates extended state observer into sliding mode controller for better transient and steady-state performance and robustness against nonlinearity, uncertainty, and time-varying disturbance. The design process of the proposed controller is composed of two main stages. Firstly, all the nonlinearity, uncertainty, and time-varying disturbance of the system are considered as the total disturbance, which is estimated by a designed high-order extended state observer (HESO). Secondly, by integrating the sliding mode control into the HESO, the composite controller is formed. Benefits from HESO's disturbance estimation and online compensation, the problem of chattering is substantially alleviated. Thanks to the sliding mode term, the closed-loop system will converge to the reference even under severe disturbances. The numerical and experimental results for a calibration-free robotic hand-eye coordination system indicate the effectiveness and robustness of the proposed approach.

High accuracy and low-cost indoor localization has received much attention for the past few years. Visible light (VL) localization, which reuses indoor LED luminaire infrastructure as artificial visual landmarks, has emerged as a promising technology. Using only lights as localization beacons, it can provide highly accurate indoor localization without reconstructing the whole environmental scenario. We propose a localization system that utilizes luminaire distribution patterns on the ceiling as visual landmarks without the costly luminaire circuits as in previous methods. For pattern detection, a unique descriptor for each luminaire cluster is derived through image processing followed by perspective rectification. Then, each pattern's ID in a lookup table is retrieved with earth mover's distance under rotation transformation. Using all individual luminaires in each pattern, our system achieves accurate and robust localization under noise and interference. Furthermore, our system can automatically generate pattern dictionaries of hundreds to thousands of unique IDs resulting in great scalability. Evaluations are conducted through real-world experiments. Compared with previous systems, the proposed method achieves superior performance in terms of detection range, localization, and orientation accuracy. Specifically, our system achieves high average three-dimensional localization accuracy of 2.47 cm and orientation accuracy of 0.87 deg under different heights for the best dictionary setting. Finally, an implementation on a mobile phone demonstrates the potential to run on resource-constrained platforms in real time.

The multiscale problem is one of the challenging topics in the object detection field. Among the current studies toward this problem, feature pyramid network (FPN) has been shown a superior performance. The core principle of FPN is a mapping for the detectable objects into hierarchy features. These features are designed with different size of receptive fields, which brings good scale invariance in the classification and localization subtasks. However, a disturbance of information dilution is generated by the assignment for the detectable objects into different features. This dilution disturbance is associated with the convolution or pooling operation during the hierarchy feature extraction process. Both of the imprecise localization information in the top features and the fuzzy classification information in the bottom ones limit the performance of FPN. Aiming at the disturbance rejection problem of information dilution, this article presents a feature compensation method. A cross compensation between the semantic information and the localization information in hierarchy features is used to enhance the vanilla FPN. Then, the evaluation metrics in an object detection task are examined. Not only the overall performance, but also the respective metrics for multiscale objects validate a considerable improvement of the FPN with a disturbance compensation mechanism. The experimental results demonstrate an effectiveness of the proposed method in this article.

Robot localization refers to the estimation of its own position based on sensor measurements. Due to the complexity of nonlinear robot systems and the disturbances in the working environment, the localization results are inevitably affected by the internal and external disturbances. Existing localization algorithms adopt a passive disturbance-rejection strategy, relying on the robustness of the algorithm itself to cope with uncertainty, and often depend on models or specific sensing configurations, which leads to limited disturbance-rejection capability and restricted application scenarios. In this paper, we propose a novel robot localization framework that can actively cope with both internal and external uncertainties of the systems. All uncertainties that affect the localization results are uniformly considered as a total disturbance.The extended state observer is designed to estimate the total disturbance, and the controller is applied to compensate the total disturbance and to improve the accuracy of localization. Compared with the traditional localization frameworks, the proposed approach is independent of specific system configurations, and can also handle arbitrary bounded disturbances,resulting in more universal, robust and effective disturbance-rejection localization strategies. Finally, the stability of the system is proved based on the Lyapunov theory. The results of simulations and a real platform-based experiment verify the effectiveness and satisfactory performance of the proposed localization framework.