Artificial Intelligence Review最新文献

Cotton aphid (Aphis gossypii Glover) severely impacts cotton yield and quality, necessitating effective detection and control measures. Traditional manual detection methods are inefficient, highlighting the need for rapid and accurate detection. In order to realize rapid and accurate cotton aphid detection, we proposed a novel attention mechanism named Bi-Enhanced Attention Mechanism (BEAM), aiming at improving the performance of the YOLOv8-s model. Furthermore, we employed a domain-adaptive transfer learning strategy by pre-training our enhanced network model on a public forestry pest dataset and fine-tuning it on a custom-built cotton aphid dataset. In this study, we evaluate our approach using the mean Average Precision (mAP) at different Intersection over Union (IoU) thresholds. Experimental results demonstrated that our approach achieved excellent performance in detecting cotton aphids. Specifically, our approach, which includes an enhanced YOLOv8-s model with a bi-enhanced attention mechanism and a domain adaptive transfer learning strategy, achieved an mAP of 58.1% at an IoU threshold range of 0.5 to 0.95 (mAP@0.5:0.95), 95.4% at an IoU threshold of 0.5 (mAP@0.5), and 64.8% at an IoU threshold of 0.75 (mAP@0.75). Compared to the baseline method, which utilizes the original YOLOv8-s model with standard training procedures, there was an improvement of 4% in mAP@0.5:0.95, 1.3% in mAP@0.5, and 8.1% in mAP@0.75. This research introduces a novel method for accurate detection of cotton aphids in the field, which is crucial for effective pest management and timely intervention.

Epilepsy is a neurological disorder characterized by abnormal neuronal discharges in the brain. As a rich source of biometric information, electroencephalography (EEG) provides favorable conditions for automated detection. Traditional algorithms and manual analysis possess solid theoretical foundations and good interpretability, however, these methods predominantly require extensive domain expertise and involve lengthy processing pipelines for complex data. The advent of artificial intelligence (AI) has facilitated the application of neural networks in the detection and prediction of epilepsy. Although such approaches heavily rely on high-quality annotated data, suffer from limited model interpretability, and involve complex training and parameter tuning, these efficient, real-time, end-to-end models still demonstrate significant potential in epilepsy analysis. This review systematically analyzes and summarizes the neural network technologies used in 341 papers published in the past three years, employing the PRISMA standard procedure. To facilitate readers’ related research, the review also summarizes the basic information of 16 publicly available datasets, common features, and metrics. Specifically, this review offers a comprehensive evaluation of diverse neural network architectures, concluding that convolutional neural networks have become a prevalent choice as classic neural networks. Furthermore, graph neural networks and transformers are experiencing a marked surge in popularity. The application of hybrid neural networks to fully extract information from EEG is also a growing trend. The review concludes with a comprehensive discussion and summary of the technical characteristics, research directions, and limitations of current methods, including patient-to-patient identification, explainable AI, dataset bias, and zone location.

Human pose estimation is a fundamental task in computer vision and robotics that involves detecting the human body joints from images or videos. It became a rapidly evolving field with applications ranging from action recognition to healthcare. This survey provides a detailed review of various methods in 2D and 3D human pose estimation for single-person and multi-person contexts in both image-based and video-based scenarios. We present a comprehensive categorization and comparison of available 2D and 3D pose datasets with an emphasis on their strengths and limitations. In addition, we also provide an overview of various evaluation metrics and loss functions commonly used to evaluate the accuracy and robustness of pose estimation models. We further discuss emerging trends, offering readers an insight into current trends in the field. We then explore key application domains where pose estimation plays an important role. The survey explains in detail about challenges in human pose estimation, including occlusion, data scarcity, privacy concerns, generalization issues, and model complexity, and suggests potential future research directions. Overall, this review aims to guide researchers in understanding current methods, datasets, and applications, while pointing out open issues and highlighting the future scope of human pose estimation.

In the current research landscape, there is a plethora of artificial intelligence methods for medical image analysis, improving diagnostic accuracy; however, AI introduces challenges related to trustworthiness and transparency of decisions. Clinicians and medical experts often find it difficult to comprehend the process by which machine learning models arrive at specific outcomes. This has the potential to hinder the ethical use of AI in a clinical setting. Explainable AI (XAI) enables clinicians to interpret and consequently improve trust for outcomes predicted by ML models. This review critically examines emerging trends in XAI applied to lung cancer modeling. Novel XAI implementations in tasks like weakly supervised lesion localization, prognostic models, and survival analysis are highlighted. Furthermore, this study explores the extend of clinician contributions in the development of XAI, the impact of interobserver variability, the evaluation and scoring of explanation maps, the adaptation of XAI methods to medical imaging, and lung-specific attributes that may influence XAI. Novel extensions to the current state-of-the-art are also discussed critically throughout this study.

In highly dynamic and high-concurrency urban traffic environments, intelligent systems must be capable of real-time perception, accurate reasoning, and agile scheduling to effectively manage complex traffic situations. However, prevailing approaches often suffer from fragmented perception, shallow reasoning, and delayed scheduling, leading to a lack of coordination among critical system modules. This deficiency significantly hinders holistic intelligent decision-making and real-time regulation under rapidly changing conditions. To address these challenges, this paper proposes a collaborative framework that integrates digital twins with metaverse-based semantic modeling. A three-layer architecture is constructed, consisting of the physical infrastructure layer, virtual twin resource layer, and traffic situation awareness layer, thereby forming a closed-loop mechanism of perception–reasoning–scheduling. The proposed system leverages the immersive semantic environment provided by the metaverse to enable contextual interpretation and intent recognition of traffic data. This semantic understanding drives the dynamic evolution and causal reasoning of digital twin entities. Based on the inferred results, a cloud–edge–end collaborative scheduling strategy is triggered to allocate resources adaptively. In addition, an interactive feedback mechanism is incorporated to support the real-time verification and continuous optimization of scheduling outcomes. Within this framework, we design a metaverse-driven Mixture-of-Experts perception network that enables multi-level semantic recognition and prediction of global traffic trends, regional congestion, and local anomalies. Furthermore, we introduce a multi-agent scheduling mechanism that combines virtualized resource mapping with structure-aware transfer strategies, thereby enhancing the generalization capacity and dynamic responsiveness of scheduling policies across heterogeneous infrastructure environments. Extensive experimental evaluations demonstrate that the proposed approach outperforms existing mainstream methods across several key metrics, including task acceptance rate, long-term revenue, congestion mitigation effectiveness, and resource utilization efficiency. These results validate the proposed framework’s superior adaptability and system-level of intelligence in complex traffic scenarios.

Non-small cell lung cancer (NSCLC), primarily consisting of lung squamous cell carcinoma (LUSC) and lung adenocarcinoma (LUAD), is a significant cause of cancer-related death globally. Accurate subtyping and identifying reliable biomarkers for NSCLC are needed to design personalized therapy; however, the molecular heterogeneity and high dimensionality of gene expression data pose significant challenges. This work proposes iPSOgs, an enhanced particle swarm optimization algorithm with a golden section search refinement strategy and an adaptive crossover-based solution generation method to improve search efficiency and convergence stability. The XGBoost classifier’s hyperparameter optimization and gene selection are done simultaneously using iPSOgs to predict NSCLC subtypes. On transcriptomic datasets (TCGA-LUAD, TCGA-LUSC, and GSE81089), the developed framework performs exceptionally well, achieving an accuracy of 0.9580, a receiver operating characteristic area under the curve of 0.9879, an F1 score of 0.9560, a recall of 0.9456, and a precision of 0.9668. The robustness of iPSOgs in managing complex optimization landscapes is further confirmed by comparison with cutting-edge metaheuristics on the CEC2017 test suite. The role of the biologically significant genes DSG3, KRT5, and SPRR2E, which were confirmed as involved in NSCLC pathogenesis by enrichment and protein-protein analysis, is further highlighted by the SHAP-based explanation. This work demonstrates the efficacy of iPSOgs as a diagnostic and biomarker discovery tool, offering a scalable and explainable solution for precision oncology in lung cancer.

Accurate estimation of a rigid object’s 6D pose and twist is a fundamental challenge for enabling autonomous systems operation and interaction with the environment. Monocular vision can mitigate the drift inherent in inertial methods and enables the estimation of non-cooperative target twist. However, the estimation accuracy of monocular vision is compromised by inherent depth ambiguity, a limitation that multi-camera systems can overcome. The lack of multi-view datasets with high-precision annotations hinders the development of robust perception algorithms. To address this, we present the first trinocular pose and twist estimation dataset for non-cooperative targets, comprising images of aircraft (7,824 for training, 4,710 for testing) and satellites (7683 for training, 4380 for testing), all manually annotated with sub-pixel-level keypoints and pose labels derived from optimization. We develop a neural network that predicts semantic keypoints for robust pose estimation. Combined with a multi-view optimization framework and twist estimation, our system achieves a mean angular velocity error of (0.1^{circ })/s and a mean linear velocity error of 0.3mm/s. Our open-source dataset and method provide a critical benchmark for future research in aerospace missions. We have open-sourced the dataset at the following https://www.kaggle.com/datasets/mingshiwuwjh/trinocular-pose-and-twist-estimation-dataset.

Structural engineering (SE) is a diverse field with numerous applications, including computational mechanics, structural simulation, and topology optimization, all governed by fundamental physical principles and typically addressed through classical numerical methods. These methods have delivered reliable and accurate solutions for forward problems within well-defined domains. However, they can become less effective when faced with high-dimensional spaces, complex geometries, irregular domains, or inverse problems with limited data. In parallel, data-driven models have gained popularity, but their dependence on large datasets and lack of physical interpretability restrict their generalization to unseen conditions. Physics-Informed Neural Networks (PINNs) have recently emerged as complementary tools that combine the strengths of numerical and data-driven approaches. By embedding governing physical laws directly into the learning process, PINNs reduce reliance on extensive datasets while improving interpretability and robustness. Although they may not yet rival classical solvers in terms of computational efficiency or accuracy for standard forward problems, PINNs offer unique advantages in scenarios where meshing is challenging, data and physics need to be integrated, or inverse problems require parameter identification and damage detection. This review provides a comprehensive overview of PINNs in SE, focusing on their theoretical framework, training strategies, computational implementations, and applications to both forward and inverse problems. The discussion highlights their advantages in accuracy, flexibility, and hybrid data-physics integration, while also outlining current limitations and future research directions to enhance their robustness and applicability for solving complex real-world SE problems.

Android is the most widely used operating system, making it a prime target for mobile malware, leading to data breaches and financial losses (e.g., Dark Herring). To address these issues, AI-based forensic tools are crucial for investigating security incidents, but their accuracy depends on high-quality mobile malware datasets. While dynamic analysis has limitations, recent research has shifted towards static analysis and AI-based methods for malware detection. However, there are three key challenges: lack of reproducibility, low dataset quality, and bias in AI datasets. This paper focuses on an overlooked bias—the incorrect API Level distribution in malware datasets. Such bias skews AI detection results, making them appear effective in tests but less applicable in real-world scenarios. To highlight the importance of dataset quality, three case studies on API Level Analysis were conducted, showing how biased datasets can distort detection results. To address this, the paper introduces methods and terms like Delayed Interception, Dataset of guaranteed quality, API Milestones, AndroBank, and Sample Unification, which aim to enhance dataset reliability and improve AI-based mobile malware detection.

Biofortified crops have gained significant attention as a sustainable solution to address malnutrition and under nutrition, particularly in developing countries. These crops are genetically enhanced to have higher levels of essential nutrients such as vitamins and minerals. It aims to improve the nutritional quality of staple foods and promote better health outcomes in populations that rely heavily on these crops. Biofortified crops play an important role in addressing global health challenges, providing a cost-effective and scalable approach to improving nutrition. But the selection of biofortified crops among various options is a complex task for decision-makers. Therefore, this paper introduces a novel approach called the feed-backward Double-hierarchy linguistic neural networks using double-hierarchy linguistic term fuzzy information to handle this issue. For this, we develop a series of weighted averaging Yager-Dombi aggregation operators and also discuss their desirable properties. The decision-making process becomes complex due to unknown weight vectors. Entropy distance measures are used to locate unknown weight vectors. The study addresses a real-world MADM problem by demonstrating that biofortified rice could potentially address vitamin A deficiency, a significant health concern in developing regions. The WASPAS approach is used to verify the proposed method, and its feasibility and efficacy are evaluated compared to other MADM techniques.