基于遥感图像的目标检测与运动目标跟踪
发布时间：2018-10-10 17:00
【摘要】：基于遥感图像的目标检测与运动目标跟踪是空间对地观测的重要研究方向，在军事监测、战场分析、交通管理等领域具有广阔的应用。针对高分辨率星载遥感图像的军事目标检测以及无人机对运动目标的稳定跟踪成为现今各国学者的研究热点之一。 本文以高分辨率遥感图像中的小型团块目标和遥感图像序列中的运动目标为研究对象，围绕目标检测和运动目标跟踪问题开展研究。文章首先总结了遥感探测基础、图像特征分析以及目标检测理论；针对高分辨率遥感图像中的小型团块目标，分析了视觉注意机制在目标检测中的作用，，基于一种新型的视觉显著性模型，综合运用背景抑制、MeanShift图像分割、目标提取等算法，有效地实现了对遥感图像中小型团块目标的检测与自动提取；其次针对遥感图像序列，分析了已有运动目标跟踪算法的原理与特点，利用卡尔曼滤波对目标运动状态的估计特性，改进原有的CamShift跟踪算法，更有效的实现了运动目标的稳定跟踪。
[Abstract]:Target detection and moving target tracking based on remote sensing image is an important research direction of space earth observation, which has wide applications in military monitoring, battlefield analysis, traffic management and so on. Military target detection based on high resolution space-borne remote sensing images and the steady tracking of moving targets by UAVs have become one of the hot research topics in many countries. In this paper, the small block target in high resolution remote sensing image and the moving target in remote sensing image sequence are taken as the research objects, and the problem of target detection and moving target tracking is studied. Firstly, the paper summarizes the basis of remote sensing detection, image feature analysis and target detection theory, and analyzes the role of visual attention mechanism in target detection for small block targets in high-resolution remote sensing images. Based on a new visual salience model, the algorithm of background suppression, MeanShift image segmentation and object extraction is used to effectively detect and automatically extract the small and medium block target of remote sensing image. Secondly, aiming at remote sensing image sequence, The principle and characteristics of the existing moving target tracking algorithms are analyzed. The original CamShift tracking algorithm is improved by using the Kalman filter to estimate the moving state of the target, and the stable tracking of moving target is realized more effectively.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751
本文编号：2262574
[Abstract]:Target detection and moving target tracking based on remote sensing image is an important research direction of space earth observation, which has wide applications in military monitoring, battlefield analysis, traffic management and so on. Military target detection based on high resolution space-borne remote sensing images and the steady tracking of moving targets by UAVs have become one of the hot research topics in many countries. In this paper, the small block target in high resolution remote sensing image and the moving target in remote sensing image sequence are taken as the research objects, and the problem of target detection and moving target tracking is studied. Firstly, the paper summarizes the basis of remote sensing detection, image feature analysis and target detection theory, and analyzes the role of visual attention mechanism in target detection for small block targets in high-resolution remote sensing images. Based on a new visual salience model, the algorithm of background suppression, MeanShift image segmentation and object extraction is used to effectively detect and automatically extract the small and medium block target of remote sensing image. Secondly, aiming at remote sensing image sequence, The principle and characteristics of the existing moving target tracking algorithms are analyzed. The original CamShift tracking algorithm is improved by using the Kalman filter to estimate the moving state of the target, and the stable tracking of moving target is realized more effectively.
【学位授予单位】：西安电子科技大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TP751
【参考文献】
相关期刊论文 前3条
1 于江德;樊孝忠;尹继豪;;基于条件随机场的中文科研论文信息抽取[J];华南理工大学学报(自然科学版);2007年09期
2 李德仁;论21世纪遥感与GIS的发展[J];武汉大学学报(信息科学版);2003年02期
3 简剑峰;尹忠海;周利华;王任享;;基于直方图不变矩的遥感影像目标匹配方法[J];西安电子科技大学学报(自然科学版);2006年04期
相关博士学位论文 前1条
1 周晖;高分辨率遥感图像的层次化分析方法[D];国防科学技术大学;2010年
本文编号：2262574
本文链接：https://www.wllwen.com/guanlilunwen/gongchengguanli/2262574.html
