Objectives: Traffic accidents cause severe social and economic impacts, demanding fast and reliable detection to minimize secondary collisions and improve emergency response. However, existing cloud-dependent detection systems often suffer from high latency and limited scalability, motivating the need for an edge-centric and consensus-free accident detection framework in IoV environments. Methods: This study presents a real-time accident detection framework tailored for Internet of Vehicles (IoV) environments. The proposed system forms an integrated IoV architecture combining on-vehicle inference, RSU-based validation, and asynchronous cloud reporting. The system integrates a lightweight ensemble of Vision Transformer (ViT) and EfficientNet models deployed on vehicle nodes to classify video frames. Accident alerts are generated only when both models agree (vehicle-level ensemble consensus), ensuring high precision. These alerts are transmitted to nearby Road Side Units (RSUs), which validate the events and broadcast safety messages without requiring inter-vehicle or inter-RSU consensus. Structured reports are also forwarded asynchronously to the cloud for long-term model retraining and risk analysis. Results: Evaluated on the CarCrash and CADP datasets, the framework achieves an F1-score of 0.96 with average decision latency below 60 ms, corresponding to an overall accuracy of 98.65% and demonstrating measurable improvement over single-model baselines. Conclusions: By combining on-vehicle inference, edge-based validation, and optional cloud integration, the proposed architecture offers both immediate responsiveness and adaptability, contrasting with traditional cloud-dependent approaches.
With the advent of the Internet of Vehicles (loV)., there is an increasing demand for real-time and accurate object detection algorithms for intelligent transportation systems. This literature review aims to provide a comprehensive analysis of the latest research in loV data processing algorithms for automatic real-time object detection. By analyzing and comparing various state-of-the-art algorithms such as Faster R-CNN., Mask R-CNN., and YOLOv4., the strengths and limitations of each approach are identified. Additionally., the challenges associated with real-world implementation are discussed and provide recommendations for future research directions. This research study highlights the importance of efficient and accurate object detection algorithms in the context of the loV and emphasizes the need for ongoing research in this field.
30 June 2026 in Sakarya University Journal of Computer and Information Sciences
The Internet of Vehicles (IoV) is witnessing escalating data traffic that demands sophisticated caching to minimize retrieval latency and sustain Quality of Service. Recent cooperative approaches, notably the Cooperative Caching Strategy with Content Request Prediction (CCCRP) framework grounded in LSTM-based request prediction, struggle to capture the complex, long-range temporal dependencies inherent to dynamic vehicular environments. This paper proposes Transformer-based CCCRP (T-CCCRP), which replaces the LSTM predictor with a Transformer model leveraging self-attention to model long-range dependencies and enable efficient parallelism. The prediction module is integrated with a reinforcement learning controller to optimize cache placement jointly across vehicles and roadside units, thereby aligning predicted popularity with resource constraints. To ensure a realistic and comprehensive evaluation, the proposed framework is assessed under both small-scale and newly introduced large-scale IoV scenarios, involving up to 200 vehicles, multiple caching nodes, and expanded content libraries. The experimental setup further adapts the input sequence length of the prediction model to reflect increased temporal complexity in large-scale environments. Simulation results demonstrate that T-CCCRP consistently outperforms the original CCCRP and conventional caching strategies (LFU and LRU) across all evaluated scales.