结合图像配准和分布式网络的轨道交通智慧监控技术  

Intelligent monitoring technology of rail transit by combining image registration and distributed network

作　　者：潘玉军 杨丽红 PAN Yujun;Yang Lihong(Wenzhou Polytechnic,Wenzhou Zhejiang 325035,China)

机构地区：[1]温州职业技术学院,浙江温州325035

出　　处：《自动化与仪器仪表》2025年第3期233-237,共5页Automation & Instrumentation

基　　金：浙江省2022年度高校国内访问学者(访问工程师)校企合作项目(FG2022049);浙江省2022年度高校国内访问学者(访问工程师)校企合作项目(FG2022050)。

摘　　要：针对传统轨道交通监控系统存在视频图像智能分析效果不佳的问题,提出设计一个结合图像配准和分布式网络的轨道交通智慧监控系统。首先,采用Spark分布式分析计算框架进行智慧监控系统构建;然后通过SURF算法对监控图像进行特征提取,并结合双向匹配和RANSAC算法进行特征匹配,以实现图像配准;最后将图像配准结果输入至系统中进行图像双光融合和目标检测,由此实现智慧监控。实验结果表明,提出的图像配准方法的mAP值高达97.32%,FPS仅为11.48 ms,明显低于其他图像配准方法,异源图像融合效果显著提升。采用Spark分布式网络构建的系统功能模块完善,满足轨道交通智慧监控图像智能分析需求,具备有效性和稳定性。In view of the poor effect of intelligent video image analysis in the traditional rail transit monitoring system,a rail transit intelligent monitoring system combining image registration and distributed network is proposed.First,Spark distributed analysis computing framework is used to build the intelligent monitoring system;then perform feature extraction through SURF algorithm and feature matching with two-way matching and RANSAC algorithm;finally input the image registration results into the system for dual-light image fusion and target detection,so as to realize intelligent monitoring.The experimental results show that the mAP value of the proposed image registration method is as high as 97.32%,and the FPS is only 11.48 ms,which is significantly lower than other image registration methods,and the heterologous image fusion effect is significantly improved.The system function module built by Spark distributed network is perfect to meet the needs of intelligent analysis of rail transit intelligent monitoring image,and has effectiveness and stability.

关 键 词：图像配准 Spark分布式网络 轨道交通 智慧监控 SURF算法 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]