| 梁敏,秦海波,覃京燕,殷绪成.AI设计下的智能驾驶场景文本识别技术[J].包装工程,2021,42(6):13-19. |
| AI设计下的智能驾驶场景文本识别技术 |
| Text Recognition Technology of Intelligent Driving Scene Based on AI Design |
| 投稿时间：2020-12-10 |
| DOI：10.19554/j.cnki.1001-3563.2021.06.003 |
| 中文关键词: 智能驾驶 交通标志 文本检测 文字识别 深度学习 |
| 英文关键词: intelligent driving traffic signs text detection character recognition deep learning |
| 基金项目:长江学者奖励项目(FRF-TP-18-010C1);国家重大专项课题(2018YFB0704301);北科大顺德研究生项目(BK19AE011) |
|
| 摘要点击次数: |
| 全文下载次数: |
| 中文摘要: |
| 目的 交通标志识别作为智能驾驶、交通系统研究中的一项重要内容，具有较大的理论价值和应用前景。尤其是文本型交通标志，其含有丰富的高层语义信息，能够提供极其丰富的道路信息。因此通过设计并实现一套新的端到端交通标志文本识别系统，达到有效缓解交通拥堵、提高道路安全的目的。方法 系统主要包括文本区域检测和文字识别两个视觉任务，并基于卷积神经网络的深度学习技术实现。首先以ResNet-50为骨干网络提取特征，并采用类FPN结构进行多层特征融合，将融合后的特征作为文本检测和识别的共享特征。文本检测定位文本区域并输出候选文本框的坐标，文字识别输出词条对应的文本字符串。结果 通过实验验证，系统在Traffic Guide Panel Dataset上取得了令人满意的结果，行识别准确率为71.08%。结论 端到端交通标志文本识别非常具有现实意义。通过卷积神经网络的深度学习技术，提出了一套端到端交通标志文本识别系统，并在开源的Traffic Guide Panel Dataset上证明了该系统的优越性。 |
| 英文摘要: |
| As an important content in the research of intelligent driving and traffic system, traffic signs recognition system has great theoretical value and application prospects. Especially, the text-based traffic signs contain rich high-level semantic information, which can provide rich road information. Therefore, according to the design and implementation of a new end-to-end text-based traffic signs recognition system, it can effectively alleviate traffic congestion and improve road safety. The system mainly includes two visual tasks:text area detection and character recognition. The system is implemented based on deep learning technology of convolutional neural network. First, resnet-50 is used as the backbone network for feature extraction, and the FPN-like structure is used for multi-layer feature fusion. The fused features are used as shared features for text detection and recognition. The text detection block locates text areas and outputs coordinates of candidate bounding boxes. The character recognition block outputs the text strings corresponding to the entry. The experimental results show that the system achieves great performance on the Traffic Guide Panel Dataset, and the F1-score is 71.08%.End-to-end text-based traffic signs recognition system is of great practical significance. According to the deep learning technology of convolutional neural network, we propose an end-to-end text-based traffic signs recognition system. Extensive experiments on publicly available dataset demonstrate the state-of-the-art performance of our method. |
|
查看全文
查看/发表评论 下载PDF阅读器 |
| 关闭 |
