Scholarlink

<p>With the rapid development of railway construction in China, railway safety has gradually attracted people’s attention and importance. The main dangers on railways are often caused by sudden intrusion of foreign objects, so railway image intelligent recognition systems are particularly important. However, traditional object detection methods are inadequate in pursuing fast and accurate detection requirements. Therefore, this study proposes an innovative intelligent detection algorithm. This algorithm combines the advantages of convolutional neural networks and single detector algorithms to identify invading foreign objects on railway tracks and is simulated and experimentally tested. The results showed that after fusing the low resolution feature maps of 75 × 75 and the high resolution feature maps of 19 × 19 and 38 × 38, the detection performance of pedestrians and trains was improved by 0.08 and 0.03, respectively. The best detection result was achieved when the model was cropped 4 times. In practical applications, the accuracy of this algorithm for detecting invading pedestrians was 99.57%. Overall, the proposed system has effectiveness, accuracy, and real-time performance in detecting intrusion targets, effectively solving the problem of slow detection speed in traditional object detection in practical applications, and making significant contributions to ensuring railway operation safety.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Traffic Safety Identification Technology Combined with Foreign Body Intrusion Intelligent Identification Technology in Railway Scene

  • 期刊论文

  • 发表日期：

    2025年10月4日

查看全文
  • Li Li

摘要

With the rapid development of railway construction in China, railway safety has gradually attracted people’s attention and importance. The main dangers on railways are often caused by sudden intrusion of foreign objects, so railway image intelligent recognition systems are particularly important. However, traditional object detection methods are inadequate in pursuing fast and accurate detection requirements. Therefore, this study proposes an innovative intelligent detection algorithm. This algorithm combines the advantages of convolutional neural networks and single detector algorithms to identify invading foreign objects on railway tracks and is simulated and experimentally tested. The results showed that after fusing the low resolution feature maps of 75 × 75 and the high resolution feature maps of 19 × 19 and 38 × 38, the detection performance of pedestrians and trains was improved by 0.08 and 0.03, respectively. The best detection result was achieved when the model was cropped 4 times. In practical applications, the accuracy of this algorithm for detecting invading pedestrians was 99.57%. Overall, the proposed system has effectiveness, accuracy, and real-time performance in detecting intrusion targets, effectively solving the problem of slow detection speed in traditional object detection in practical applications, and making significant contributions to ensuring railway operation safety.