Aug 2026· International Journal of Science and Research Archive· 0 citations
TL;DR
An effective real-time traffic accident detection framework based on YOLOv8 that can be implemented in intelligent transportation systems, traffic surveillance platforms, and advanced driver assistance applications is proposed.
Abstract
Traffic accidents remain a crucial problem for intelligent transportation systems (ITS), and the ability to detect them quickly and efficiently is one of the keys to ensuring an appropriate emergency response and improving road safety. Traditional accident monitoring methods often suffer from poor detection accuracy, high computational costs, and reduced robustness when dealing with complex traffic conditions. This paper presents a novel enhanced YOLOv8-based framework for real-time traffic accident detection capable of operating in dynamic road environments.
The proposed framework uses YOLOv8 as the base detection model, which provides powerful feature extraction capabilities, an anchor-free detection mechanism, and a favorable trade-off between accuracy and inference speed. To address the challenges associated with detecting accident-related objects, such as occlusion, illumination variations, and complex traffic backgrounds, the proposed framework incorporates feature augmentation to enhance feature representation. Multi-scale features are fused at different scales to improve local details, thereby enabling the extraction of critical regional features.
Experimental results show that the proposed framework achieves competitive detection performance, with a precision of 94.3%, a recall of 92.7%, and an F1-score of 93.5%, as well as a mAP@0.5 of 96.1%. Furthermore, the proposed framework achieves a real-time inference speed of 58 FPS on the COCO test-dev 2017 dataset, compared with its own smaller YOLOv4-Nano model with an input size of 416 × 416 pixels. Qualitative results further demonstrate the ability of the proposed approach to localize traffic objects in both successful and direct accident scenarios.
In this paper, we propose an effective real-time traffic accident detection framework based on YOLOv8 that can be implemented in intelligent transportation systems, traffic surveillance platforms, and advanced driver assistance applications.
In recent years, traffic accidents on highways have occurred frequently. An accident threatens lives and property. It also disrupts traffic. Current accident scene recognition faces challenges. Accuracy is often suboptimal. Deploying detection models in real-world scenarios is difficult due to high computational complexity and insufficient inference speed, rather than parameter size alone. These scenarios have limited computational resources. High computational complexity and low inference efficiency cause the difficulty, while model parameter size is not the only bottleneck. To address these issues, a lightweight highway accident scene recognition algorithm based on an improved YOLOv8n is proposed. This algorithm is designed with potential deployment in fixed monitoring devices installed at accident-prone highway sections in mind. It enables automatic accident detection through real-time video analysis and could promptly send alerts to traffic management centers, potentially facilitating rapid dispatch of rescue resources and traffic control, thereby offering a pathway to enhance highway safety management and emergency response efficiency. However, we note that field deployment experiments and hardware-level latency tests are not included in this study, and thus practical applicability claims remain to be validated in future work. First, MobileViT, a lightweight CNN, enhances feature extraction for highway accident targets, showing improved robustness under simulated adverse weather conditions (e.g. blurring and contrast reduction that mimic fog and rain). However, we explicitly note that these are image-space simulations; the model’s generalization to authentic foggy or rainy environments has not been tested in this study and remains a critical direction for future investigation. Second, Ghostnet and VanillaNet are added to the model’s neck network, reducing computational complexity and improving inference speed while maintaining detection accuracy and keeping parameter growth minimal in extreme weather. Finally, SlideLoss replaces the original Loss function to address sample imbalance and improve detection of complex targets. Ablation and comparative experiments were conducted using the DADA and Car Crash Dataset datasets. The proposed algorithm achieves higher average precision than traditional methods. The improved model’s average precision increased by 2.3%, from 92.8% to 95.1%, while also adhering to lightweight design principles, with computational complexity reduced from 8.2 to 7.2 and the number of parameters decreased from 3.0 M to 2.9 M. These findings confirm the algorithm’s superiority and the effectiveness of its improvements.
Dengcong Mu, Zitang Wei, Zheng Li et al.· Engineering Research Express· 0 citations
Smart city transportation has become an important part of improving road safety and traffic management. This project focuses on detecting traffic accidents using a deep learning ensemble approach that combines I3DConvLSTM2D with RGB and optical flow information. By analyzing both the appearance of vehicles and their movement, the system can identify accident events more accurately than traditional methods. It is designed to work in real time, making it suitable for surveillance cameras and smart city environments. The proposed model also addresses challenges such as limited training data and different road conditions, helping improve its reliability in practical situations. Early accident detection allows emergency services to respond quickly, reducing the impact of road accidents and improving public safety. Overall, this system demonstrates how artificial intelligence can support modern transportation by providing faster, more accurate, and efficient accident detection while contributing to safer roads and better traffic monitoring.
CHENCHETI INDHU, Dr.M.Ramesh· International Journal of Eng...· 0 citations
Road traffic accidents require rapid detection and timely warning dissemination to reduce secondary collisions and improve emergency response. This paper presents an efficient roadside-unit-based computer vision framework for real-time accident recognition integrated with Vehicle-to-Everything communication. The proposed architecture combines lightweight convolutional spatial feature extraction, optical-flow-based motion analysis, temporal score aggregation, and low-latency V2X alert generation within an edge-deployable RSU pipeline. A formal system model is introduced to describe accident likelihood estimation, detection decision making, latency constraints, and communication reliability requirements. The experimental evaluation was conducted on a traffic video dataset containing 3,300 clips with normal traffic, single-vehicle accidents, multi-vehicle collisions, and non-accident anomalies. The proposed hybrid framework achieved 95.8 % accuracy, 95.0 % precision, 94.6 % recall, and 94.8 % F1-score, outperforming CNN-only and motion-only baselines. The average processing latency was approximately 40 ms per frame, indicating the feasibility of real-time operation on RSU-grade embedded hardware. These results show that the proposed framework can provide accurate and timely accident detection while supporting rapid V2X warning dissemination for next-generation intelligent transportation systems.
Danish Ather, M. Talipov· Mathematical Models in Engin...· 0 citations
To address the challenges of construction vehicle detection caused by complex illumination and severe occlusion in transportation infrastructure scenarios, as well as the requirements for emergency rescue under sudden hazards, this paper proposes a detection system based on a multi-model decision fusion architecture. The system adopts a “coarse-screening and fine-inspection” strategy, utilizing YOLOv8 for rapid global screening to ensure real-time performance. For regions with low confidence, an improved DPFL-RetinaNet is introduced to perform secondary refined detection. Concurrently, this paper proposes the Decoupled Polynomial Focal Loss based on Taylor series expansion, which enhances the mining capability for weak targets through gradient reshaping. Experimental results demonstrate that the proposed system achieves an AP50 of 87.1% on the TCVD dataset, representing an improvement of nearly 30 percentage points over the standalone YOLOv8. This provides an efficient visual perception solution for construction safety and emergency rescue in complex environments, effectively balancing detection precision and inference speed. The proposed Confidence-Guided Cascade Decision System (CG-CDS) effectively addresses the inherent trade-off between computational efficiency and detection accuracy. By dynamically routing challenging samples to a specialized fine-grained detector, the system achieves robust performance across diverse environmental conditions including low illumination, severe occlusion, and complex background clutter. Extensive ablation studies confirm that each component of the cascaded architecture contributes positively to the overall detection performance, with the proposed Decoupled Polynomial Focal Loss (DPFL) providing the most significant improvement for weak target mining.
IYOLO, an enhanced YOLOv8-based framework for simultaneous detection and classification of vehicles, drivers, and passengers on highways, aiming to distinguish drivers from passengers and establish one-to-one vehicle-driver associations is proposed.
Yang Zhang, Peihua Lv, Hongjin Ren et al.· International Conference on...· 0 citations
With the increasing number of vehicles, urbanization and the constant rise in road usage, traffic violations have become hugely problematic in today's transportation situations. Some of the most common dangerous driving behaviors that lead to road accidents and traffic delays are as follows: Not wearing a helmet, running a red light, and breaking lanes, breaking seatbelt, using a cell phone while driving and triple riding. Maintaining consistent observation, precise detection, scalability, and speedy identification of traffic infractions in complex road situations are all challenges faced by current traffic violation monitoring methods. Factors such as high traffic levels, uneven lighting, environmental interference, and requiring human supervision limit the effectiveness of current monitoring methods. Therefore, it becomes essential to have a sophisticated automated system that can efficiently do real-time traffic infraction analysis. The proposed study utilizes a novel Traffic violations identification method based on YOLOv8 to detect many traffic violations in the surveillance photos and videos. To achieve the system execution, a Traffic Rule-net Dataset was developed from a set of traffic data collected from different scenarios of city transport, highways and crossroads. The quality of the features and the robustness of the suggested model were enhanced using a number of data pre-processing techniques, including normalization, image size modification, data augmentation, and filtering. The new framework was applicable in all environmental conditions, allowing for efficient object localization and classification of different breaches. From the experimental evaluation it is clear that there was a reduction in false detection, performance, and detection efficiency. Infrancements of the traffic rules may be easily and efficiently detected for traffic control through the use of intelligent monitoring.
Selvam L, G. Aninthitha, P. M et al.· 2026 7th International Confe...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
