With the development of artificial intelligence technology and the popularity of intelligent production projects, intelligent inspection systems have gradually become a hot topic in the industrial field. As a fundamental problem in the field of computer vision, how to achieve object detection in the industry while taking into account the accuracy and real-time detection is an important challenge in the development of intelligent detection systems. The detection of defects on steel surfaces is an important application of object detection in the industry. Correct and fast detection of surface defects can greatly improve productivity and product quality. To this end, this paper introduces the MSFT-YOLO model, which is improved based on the one-stage detector. The MSFT-YOLO model is proposed for the industrial scenario in which the image background interference is great, the defect category is easily confused, the defect scale changes a great deal, and the detection results of small defects are poor. By adding the TRANS module, which is designed based on Transformer, to the backbone and detection headers, the features can be combined with global information. The fusion of features at different scales by combining multi-scale feature fusion structures enhances the dynamic adjustment of the detector to objects at different scales. To further improve the performance of MSFT-YOLO, we also introduce plenty of effective strategies, such as data augmentation and multi-step training methods. The test results on the NEU-DET dataset show that MSPF-YOLO can achieve real-time detection, and the average detection accuracy of MSFT-YOLO is 75.2, improving about 7% compared to the baseline model (YOLOv5) and 18% compared to Faster R-CNN, which is advantageous and inspiring.

Feature selection and dimensionality reduction are important for the performance of wind turbine condition monitoring models using supervisory control and data acquisition (SCADA) data. In this paper, an improved random forest algorithm, namely Feature Simplification Random Forest (FS_RF), is proposed, which is capable of identifying features closely correlated with wind turbine working conditions. The Euclidian distances are employed to distinguish the weight of the same feature among different samples, and its importance is measured by means of the random forest algorithm. The selected features are finally verified by a two-layer gated recurrent unit (GRU) neural network facilitating condition monitoring. The experimental results demonstrate the capacity and effectiveness of the proposed method for wind turbine condition monitoring.

Immersive systems, which offer immersive interactive experiences by tightly coupling the real and virtual worlds, encompassing Augmented Reality (AR) and Virtual Reality (VR), have undergone rapid development in recent years. The inherent latency issue of immersive systems poses considerable challenges for AR/VR applications that are highly sensitive to delay. To address this challenge, we propose utilizing edge collaboration as an efficient means to schedule AR/VR requests, owing to the inherent ability of edge nodes to seamlessly coordinate within the same geographical region and optimally allocate request resources to minimize latency. However, when it comes to edge nodes across different regions, a significant hurdle emerges. These nodes often lack real-time knowledge about cached resources across other regions, making cross-regional edge collaboration for AR/VR request scheduling an extremely challenging problem. We formally define and prove the NP-hardness of this problem. To tackle this complexity, we introduce a Federated Digital Twin Model Construction (FedDT), which establishes a global digital twin network connecting different edge nodes. This network simulates and reflects the resource status and availability of each edge node in real time. Subsequently, we devise an extended Dijkstra algorithm to determine the shortest path between AR/VR requests and edge nodes, taking into account current network conditions to optimize service delay. Both theoretical analysis and experimental results confirm that compared with baseline methods, our proposed solution consistently achieves an average delay reduction of 36.78% across various scenarios, effectively demonstrating its superior performance.

In smart cities, Multi-Camera Multi-Target pedestrian tracking and Re-identification (MCMT-ReID) is essential for effective surveillance, particularly in real-time scenarios, as it demands significant computational resources. Current edge-cloud collaboration methods encounter issues such as high latency and potential data leakage due to the physical distance between cloud servers and cameras. To address these issues, we propose a novel Edge-to-Edge Collaboration (E2EC) system that fully utilizes collaboration between heterogeneous edge devices. E2EC partitions the MCMT-ReID task into two modular applications: Tracking and Re-identification (ReID), and employs a customized Kafka communication protocol to optimize data exchange efficiency. Moreover, E2EC dynamically orchestrates intermediate inference flows and transmits features instead of pedestrian detection frames to avoid data leakage. To enhance ReID accuracy, we introduce a real-time ReID Loop Confirmation (ReLC) algorithm, which continuously validates identities to boost reliability and accuracy. E2EC has been deployed and tested in a real-world campus environment to validate its effectiveness. Experimental results demonstrate that E2EC enhances the Rank-1 accuracy and mAP of pedestrian ReID by 36.88% and 46.00%, respectively. Furthermore, it achieves an increase of about 6.35%-12.66% in throughput and reduces latency by 35.01%-57.83% compared to baselines, ensuring real-time performance under dynamic workloads.

Real-time person Re-Identification (ReID) services in video surveillance remain challenging due to real-time requirements and limited generalization of supervised models. Traditional supervised methods typically require extensive labeled training data and are confined to closed identities, struggling to generalize to unseen individuals or new environments. To overcome these limitations, recent work leverages large Vision-Language Models (VLMs) for zero-shot ReID, utilizing their rich visual-textual representations to identify pedestrians without prior training. However, directly deploying such VLMs in the cloud inherently introduces high latency, further hindering real-time performance. To tackle the above challenges, we propose VL-ZSReID, a novel edge-collaborative inference framework that deploys VLMs on resource-constrained edge devices to enable zero-shot ReID in complex multi-camera campus surveillance environments. The framework partitions the ReID task into a Tracking Service and a VLM Service, which are coordinated by a Cache-assisted Heuristic Edge Scheduling Strategy (CHESS) to reduce latency and improve throughput across heterogeneous edge devices. To enable practical deployment of VLMs on resource-constrained edge devices, VL-ZSReID adopts several system-level design choices. It incorporates a Tracking ID Discriminator (TID) as an intelligent caching module, an Attribute Caption Generation (ACG) mechanism that produces standardized feature descriptions, and a Temporal-Camera Feature Generator (TCFG) that leverages spatiotemporal context to improve accuracy. Implemented and evaluated in a real campus deployment, VL-ZSReID achieves 19.98%–195.47% higher throughput and 38.00%–82.27% lower latency than baseline methods, while improving Rank-1 and mAP by 2.44%–44.68% and 4.11%–60.89%, respectively.