卫星视频运动车辆检测方法的评价分析与优化  

Evaluation and optimization of moving vehicle detection from satellite videos

作　　者：王璐 陈仁喜[2] WANG Lu;CHEN Renxi(The First Geological Brigade of Jiangsu Geological Bureau,Nanjing 210041,China;School of Geography and Remote Sensing,Hohai University,Nanjing 211100,China)

机构地区：[1]江苏省地质局第一地质大队,南京210041 [2]河海大学地理与遥感学院,南京211100

出　　处：《智能计算机与应用》2025年第3期100-105,共6页Intelligent Computer and Applications

基　　金：中国科学院太空应用重点实验室开放基金(LSU-KFJJ-2018-10)。

摘　　要：近年来视频卫星的成功研发,使对地表动态目标的实时性监测成为可能。为将视频卫星运用于智能交通领域,首先需实现卫星视频中车辆目标的高效检测,该项技术是获取一切交通信息的前提。然而,卫星拍摄的视频数据与地面监控视频差异悬殊,现有检测算法难以全面、且准确地检测出卫星视频中车辆目标。为此,本研究在对现有检测算法适用性进行评价的基础上,结合视频影像特点,引入特征背景建模法,通过构建非感兴趣区掩膜,剔除大量误检目标,以获取更加准确全面的车辆检测结果。本研究以吉林一号卫星视频作为实验数据,经实验对比,在兼顾查全率的情况下,算法优化后获得的检测结果在检测精准度上得到显著提升。In recent years,the successful development of video satellites makes it possible to monitor dynamic targets on the ground in real time.In order to apply the video satellite to the field of intelligent transportation,it is necessary to achieve the efficient detection of vehicle targets in the satellite video,which is the premise of obtaining all traffic information.However,due to the great difference between the video data captured by satellite and the ground surveillance video,it is difficult for the existing detection algorithms to detect the vehicle targets in the satellite video accurately and comprehensively.Therefore,based on the applicability evaluation of the existing detection algorithms and considering the characteristics of video images,this paper introduces the eigenback ground modeling method.By constructing the non-ROI mask,a large number of false detection targets are eliminated to obtain more accurate and comprehensive vehicle detection results.In this study,Jilin-1 satellite video is used as experimental data.Through experimental comparison,while taking into account the recall ratio,the detection accuracy has been significantly improved in the detection results obtained after optimization of the algorithm.

关 键 词：视频卫星 遥感影像 目标检测 吉林一号 

分 类 号：TP751[自动化与计算机技术—检测技术与自动化装置] P237[自动化与计算机技术—控制科学与工程]