基于复合信号特征的崩塌落石监测方法

A Composite-Feature-Based Method for Rockfall Hazard Surveillance

行车限界内的崩塌落石灾害监测是保障山区铁路行车安全的重点研究课题.若干基于单信号特征的监测方法已经投入应用,其固有的误报率和漏报率超出了铁路运输所能容忍的限度.考虑到崩塌落石会产生侵限目标和冲击波等多种信号特征,提出了一种基于复合信号特征的监测方法,对信号源不同的多种信号特征进行提取,同时要求这些特征必须具备高度的时空域相关性.理论分析和各山区路局的实践表明,此方法可以使监测系统的误报率和漏报率同时降低到若干数量级以下.

Rockfall hazard surveillance inside traffic clearance is an important research issue for railway transportation safety in mountain area.Some single-feature-based detecting methods have been proposed and practiced,whose intrinsic False-positive-ratio （FPR） and False-negative-ratio （FNR） are worse than railway transportation can tolerate.Considering the fact that rockfall on railway line produces various signal features including object appearance and impulse wave,a composite-feature-based detecting method which imposes spatiotemporal constraint on features extracted from multiple signal sources is proposed.Theoretical analysis and practice in mountain railway show that FPR/FNR with several orders lower can be obtained simultaneously.