eismic method is one of the main tools of assessing fault activity. There are Pwave reflection and shearwave splitting methods when active fault is investigated using seismic methods. The Pwave reflection method is ac...eismic method is one of the main tools of assessing fault activity. There are Pwave reflection and shearwave splitting methods when active fault is investigated using seismic methods. The Pwave reflection method is according to the fact that whether bedrock fault extends up to Quaternary strata Q3 or not, while shearwave splitting method is according to anisotropy of Quaternary strata Q3 determined by the shearwave splitting in assessing activity of bedrock fault. The Pwave reflection method can be used under the conditions of large scale bedrock fault up to Quaternary strata, whereas the shear wave splitting method can be used.展开更多
当前高速铁路接触网参数检测中,存在开口销体积小、分布分散、故障缺陷识别困难,过度依赖综合检测车等问题。本文提出一种采用无人机航拍,结合图像分割与识别技术的基于更快的区域卷积神经网络(Faster R-CNN)算法实现图像处理和优化,进...当前高速铁路接触网参数检测中,存在开口销体积小、分布分散、故障缺陷识别困难,过度依赖综合检测车等问题。本文提出一种采用无人机航拍,结合图像分割与识别技术的基于更快的区域卷积神经网络(Faster R-CNN)算法实现图像处理和优化,进而对开口销缺陷进行检测识别的方法,有效地提升开口销缺陷识别准确率和有效性。测试结果表明,采用基于Faster R-CNN算法的无人机高速铁路接触网开口销缺陷检测方法的开口销图像缺陷识别准确率可达到98%以上,平均精度约90%,接受者操作特征曲线下的面积(area under curve,AUC)大于0.98。该算法通过软件开发工具包(software development kit,SDK)嵌入到无人机,实现接触网开口销自动巡检、智能识别,为现场作业提供智能化检测设备,提升接触网的智能化检测手段,保障高速铁路安全运行。展开更多
文摘eismic method is one of the main tools of assessing fault activity. There are Pwave reflection and shearwave splitting methods when active fault is investigated using seismic methods. The Pwave reflection method is according to the fact that whether bedrock fault extends up to Quaternary strata Q3 or not, while shearwave splitting method is according to anisotropy of Quaternary strata Q3 determined by the shearwave splitting in assessing activity of bedrock fault. The Pwave reflection method can be used under the conditions of large scale bedrock fault up to Quaternary strata, whereas the shear wave splitting method can be used.
文摘当前高速铁路接触网参数检测中,存在开口销体积小、分布分散、故障缺陷识别困难,过度依赖综合检测车等问题。本文提出一种采用无人机航拍,结合图像分割与识别技术的基于更快的区域卷积神经网络(Faster R-CNN)算法实现图像处理和优化,进而对开口销缺陷进行检测识别的方法,有效地提升开口销缺陷识别准确率和有效性。测试结果表明,采用基于Faster R-CNN算法的无人机高速铁路接触网开口销缺陷检测方法的开口销图像缺陷识别准确率可达到98%以上,平均精度约90%,接受者操作特征曲线下的面积(area under curve,AUC)大于0.98。该算法通过软件开发工具包(software development kit,SDK)嵌入到无人机,实现接触网开口销自动巡检、智能识别,为现场作业提供智能化检测设备,提升接触网的智能化检测手段,保障高速铁路安全运行。
