Research on Traffic Sign Recognition Based on Improved YOLOv11n.
Saved in:
| Title: | Research on Traffic Sign Recognition Based on Improved YOLOv11n. |
|---|---|
| Authors: | Guan, Zhiwei1 2366116174@qq.com, Xu, Yang xuyang_1981@aliyun.com |
| Source: | Engineering Letters. May2026, Vol. 34 Issue 5, p2026-2041. 16p. |
| Subjects: | Traffic signs & signals, Object recognition (Computer vision), Convolutional neural networks, Deep learning, Computer vision, Image processing, Feature extraction |
| Abstract: | To address the problems of insufficient recognition accuracy, missed detection, or a high false detection rate of traffic signs under complex environmental conditions, this paper proposes an improved traffic sign recognition algorithm, SLS-YOLOv11n, based on YOLOv11n (You Only Look Once Version 11). Firstly, SPD-Conv is used to replace some traditional convolutions, and different types of convolutions are employed to distribute the computational cost and capture more diverse feature information. At the same time, both shallow details and deep semantics are considered to eliminate the problem of detail loss caused by step convolution and to improve the ability to detect small targets. Then, the lightweight Edge-Gaussian Driven Network Module (LFEM) is integrated into the original C3k2 module to enhance the feature representation. Especially when dealing with low-quality images, it has good robustness while maintaining computational efficiency. Finally, the feedforward neural network SEFFN (Spectral Enhanced FFN) with channel attention mechanism enhances the ability of the model to emphasize key channel features and further improves the accuracy of model detection. This paper conducts experiments on two public transportation sign datasets, TT100K and CCTSDB2021. In this paper, the experimental verification is carried out on the two public transport sign data sets of TT100K and CCTSDB2021. The experimental results show that on the TT100K dataset, the Precision, Recall, and mAP0.5 (mean Average Precision) of the original model are 70.3, 65.8, and 71.8, respectively. In contrast, the improved model increases them by 4.1, 1.9, and 3.4 percentage points, respectively. On the CCTSDB2021 dataset, the Precision, Recall, and mAP0.5 of the original model were 84.2, 64.0, and 72.8, respectively, while the improved model increased by 0.2, 3.2, and 2.9 percentage points, respectively. Compared with the original model, the comprehensive performance of the enhanced model is improved and has good performance. [ABSTRACT FROM AUTHOR] |
| Copyright of Engineering Letters is the property of International Association of Engineers (IAENG) and its content may not be copied or emailed to multiple sites without the copyright holder's express written permission. Additionally, content may not be used with any artificial intelligence tools or machine learning technologies. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.) | |
| Database: | Engineering Source |
Be the first to leave a comment!
