Augmented Reality (AR) systems critically depend on accurate, temporally stable, and computationally efficient object tracking to maintain geometric alignment and perceptual coherence between virtual content and the physical world. As AR technologies transition from controlled laboratory prototypes to large-scale deployment in industrial, medical, and consumer scenarios, tracking must operate robustly in complex real-world environments characterized by dynamic objects, occlusions, illumination changes, fast motion, and strict computational constraints. Traditional geometry-driven tracking pipelines often degrade under such conditions, motivating increased adoption of deep learning based approaches. This survey provides a comprehensive review of deep learning based object tracking for AR in complex real-world scenes, with particular emphasis on system-level considerations. Object tracking is treated as a central perception primitive that underpins stable AR experiences and interacts tightly with modules such as visual simultaneous localization and mapping, depth and geometry estimation, and semantic scene understanding. In contrast to prior surveys that emphasize algorithmic accuracy in isolation, we explicitly analyze AR-specific constraints including real-time latency, temporal stability, energy efficiency, long-term robustness, and deployment on mobile and wearable platforms. We review major tracking paradigms, representative datasets and benchmarks, AR-centric evaluation criteria, and common failure modes observed in practice, and we outline future research directions toward scalable, reliable, and trustworthy AR tracking systems.

This paper systematically studies the non-invasive continuous blood pressure estimation and cardiovascular risk assessment methods based on photoplethysmography (PPG) signals. A complete signal processing flow was constructed using the PhysioNet public dataset, including preprocessing, feature extraction, model construction and validation. The experimental results show that the extraction of heart rate and respiratory rate based on PPG has high accuracy, preliminarily verifying the correlation between the morphological characteristics of PPG and blood pressure parameters. The research further proposed a hierarchical processing architecture to balance the computing efficiency and monitoring depth of wearable devices, providing a reference for the engineering implementation of health monitoring systems.

Gearbox lubricating oil oxidation degradation severely impairs the operational stability of wind turbines and increases maintenance costs. Conventional detection methods (e.g., Rotating Pressure Vessel Oxidation Test (RPVOT), Pressure Differential Scanning Calorimetry (PDSC)) have high instrument dependence and long testing cycles, and may not meet on-site rapid detection demands. This study took in-service industrial gear oil samples as the research object, selected five physicochemical indices as input variables, and systematically optimizes data preprocessing (Standard Normal Variate (SNV)), feature extraction (Principal Component Analysis (PCA)), and machine learning algorithms (Back Propagation Neural Network (BPNN)/Support Vector Machine (SVM)/Random Forest (RF)). The proposed SNV-PCA-BP hybrid model achieved excellent predictive performance with a R2 of 0.9960 and Residual Prediction Deviation (RPD) of 6.1247, which is 360 times more efficient than traditional methods. This model provides a low-cost and reliable technical support for the predictive maintenance of wind turbine gearboxes.

Faced with the technical bottlenecks of traditional single-sensor methods in dynamic pipeline environments, this study proposes a hierarchical progressive intelligent framework for pipeline robot navigation and defect recognition based on laser-vision-inertial multi-sensor fusion. An innovative Unscented Kalman Filter (UKF)-based nonlinear optimization framework unifies state vectors of navigation and defect recognition to achieve collaborative optimization of pose estimation and feature extraction. The bottom layer realizes multi-source data spatiotemporal alignment via extended Kalman filter; the middle layer completes environmental modeling using improved ICP and visual SLAM（Simultaneous Localization and Mapping）; the top layer adopts lightweight CNN for defect identification. Experiments show that the system achieves centimeter-level positioning accuracy and 92.7% defect recognition accuracy in complex scenes, with cumulative positioning error controlled within ±8 cm and robustness improved by over 40%. It is universal for 200–800 mm pipe diameters, reducing small-diameter navigation drift by 42% and reaching 98.3% detection coverage. Engineering tests verify 4.2× higher inspection efficiency and 65% lower maintenance costs than manual operations. With modular hardware and open software, the system adapts to oil and gas pipelines, utility tunnels and drainage systems, and its cloud-edge knowledge base supports rapid scenario adaptation via transfer learning, providing reusable methodological support for intelligent perception systems.

This study examines the application of AI models in dynamic talent adaptation between industry and education, using data from the Digital Genealogy Cloud Platform as a research sample. It analyzes current challenges in industry-education talent alignment and elaborates on the pivotal role of AI models. Through in-depth mining and analysis of platform data, the research demonstrates how AI models leverage big data, artificial intelligence, and knowledge graphs to achieve precise digital integration across industrial chains, talent pipelines, and educational systems. The findings reveal that AI models not only assist vocational colleges in optimizing program structures and innovating talent cultivation models, but also facilitate deep collaboration between schools and enterprises. This enhances the alignment between talent development and industrial demands, providing robust support for the digital transformation and high-quality development of vocational education. Additionally, it offers new perspectives and practical solutions to address fundamental issues in industry-education integration.

This paper focuses on specific products in consumer finance and explores the innovative application of artificial intelligence algorithms in credit scoring. By analyzing the technical foundation and current situation, this paper expounds the core technological innovations such as deep learning, tree model integration, and graph neural networks, and solves the problems of data fusion, model optimization, and interpretability. Empirical research shows that artificial intelligence algorithms outperform traditional models in multiple indicators.