Lightweight ViT with Multiscale Feature Fusion for Driving Risk Rating Warning System
Hao Tang,Xixi Xu,Haiyang Xu,Shuang Liu,Jie Ji,Chengqun Qiu,Yujie Shen
DOI: https://doi.org/10.1002/adts.202400586
2024-08-16
Advanced Theory and Simulations
Abstract:This study introduces the RepBF‐YOLOv8 algorithm, enhancing dynamic object detection for autonomous driving. Employing the lightweight RepViT and advanced Bi‐FPN, offering refined risk assessment through efficient multi‐scale feature fusion and rapid risk categorization, significantly improving precision and recall while maintaining real‐time performance. The study validates the model across various traffic scenarios, demonstrating robustness, and adaptability. Addressing the issue of inadequate dynamic object detection accuracy in current road driving warning systems, this study proposes the RepBF‐YOLOv8 detection algorithm aimed at efficient risk identification. The backbone network of YOLOv8n is replaced with the lightweight RepViT architecture, which is more suitable for visual tasks. This replacement simplifies the traditional structure, reduces the complexity of the backbone network, maximizes performance enhancement, and minimizes latency. Additionally, the FPN in the neck section is upgraded to Bi‐FPN, which reduces nodes and span connections and incorporates rapid normalization to achieve fast multi‐scale feature fusion. For risk grading, the algorithm infers distances and collision times, categorizing detected objects into high, medium, and low‐risk levels, and uses different colors to warn the driver. Comparative experimental results show that the optimized algorithm improves Precision by 1.7%, Recall by 2.3%, mAP@0.5 by 1.53%, and mAP@0.5:0.95 by 2.91%. In road tests, the risk warning system achieves a frame detection rate ranging from a minimum of 38.4 fps to a maximum of 59.0 fps. The detection confidence for various objects remains above 0.71, reaching as high as 0.98. Specifically, the "Car" confidence ranges from 0.81 to 0.98, demonstrating the accuracy and robustness of vehicle risk detection.
multidisciplinary sciences