基于视觉注意机制的UWB SAR叶簇隐蔽目标变化检测技术研究
发布时间：2019-05-27 16:48
【摘要】：低频超宽带合成孔径雷达(Uitra Wide Band Synthetic Aperture Radar,UWB SAR)因其强叶簇穿透能力,逐步在民事、军事领域受到广泛应用。然而,随着UWB SAR系统分辨率提高,应用范围的不断扩大,对于UWB SAR叶簇隐蔽目标快速、精准检测的需求日趋强烈。因此本文从人类视觉系统的视觉注意机制入手,分析并建立适用于低频UWB SAR图像的视觉注意模型,提出了一种无监督、全自动、高效率的低频UWB SAR叶簇隐蔽目标变化检测方法,取得了良好的检测效果。论文工作及创新点主要包括:1、视觉注意机制与图像处理相结合的可行性分析。本文依据人类视觉系统的现有资料,从生物学角度入手,研究人类视觉系统工作机理,总结其信息处理的主要特点;探讨将视觉注意机制应用于低频UWB SAR图像处理过程的可行性,并构建适用于低频UWB SAR图像处理的视觉注意模型,为后期结合视觉注意机制展开低频UWB SAR变化检测技术的研究提供理论基础。2、低频UWB SAR图像数据预处理,即图像可视化技术研究。首先结合人眼视觉特性,确定低频UWB SAR图像可视化的视觉依据,分析低频UWB SAR图像内在的数据统计特性和灰度分布模型,并针对低频UWB SAR图像的特点,提出符合视觉特性的低频UWB SAR快速可视化算法。在同等条件下该方法与传统的对数映射法、线性映射法相比,所得结果在等效视数、处理耗时等指标的考核中表现优异,更适于人眼判读,且能够满足低频UWB SAR系统实时处理等方面的应用需求。3、基于视觉注意机制展开多时相UWB SAR叶簇隐蔽目标变化检测技术研究。首先从视觉分析角度出发,模拟人类视觉系统,制构建适用于图像分析的高斯金字塔模型,根据人类视觉系统的生理结构与认知特点,提出基于视觉注意机制的UWB SAR叶簇隐蔽目标变化检测算法,该算法将复杂图像简化为视觉焦点集合,并利用图像局部邻域信息和目标的空间相关特性对视觉注意焦点进行分层筛选,以提高变化检测精度。与广泛应用的基于CFAR的变化检测技术相比,同等情况下,基于视觉注意机制的UWB SAR叶簇隐蔽目标检测算法不但能有效弥补CFAR检测方法存在的漏检问题,而且处理耗时量仅为CFAR检测的12%,因而在实际应用中具有重大的应用前景。
[Abstract]:Low frequency ultra-broadband synthetic aperture radar (Uitra Wide Band Synthetic Aperture Radar,UWB SAR) has been widely used in civil and military fields because of its strong blade cluster penetration ability. However, with the improvement of the resolution of UWB SAR system and the continuous expansion of its application range, the demand for fast hidden targets and accurate detection of UWB SAR leaf clusters is becoming more and more strong. Therefore, starting with the visual attention mechanism of human visual system, this paper analyzes and establishes a visual attention model suitable for low frequency UWB SAR images, and proposes an unsupervised, fully automatic and efficient method for detecting the change of hidden targets in low frequency UWB SAR leaf clusters. Good detection results have been obtained. The main work and innovations of this paper are as follows: 1. The feasibility analysis of the combination of visual attention mechanism and image processing. Based on the existing data of human visual system, this paper studies the working mechanism of human visual system from the biological point of view, and summarizes the main characteristics of its information processing. This paper discusses the feasibility of applying visual attention mechanism to low frequency UWB SAR image processing, and constructs a visual attention model suitable for low frequency UWB SAR image processing. It provides a theoretical basis for the later research of low frequency UWB SAR change detection technology combined with visual attention mechanism. 2, low frequency UWB SAR image data preprocessing, that is, image visualization technology research. Firstly, combined with the human visual characteristics, the visual basis of low frequency UWB SAR image visualization is determined, and the inherent data statistical characteristics and gray distribution model of low frequency UWB SAR image are analyzed, and the characteristics of low frequency UWB SAR image are analyzed. A fast visualization algorithm for low frequency UWB SAR is proposed, which accords with visual characteristics. Under the same conditions, compared with the traditional logarithmic mapping method and linear mapping method, the results obtained by this method have excellent performance in the assessment of equivalent visual number and processing time, and are more suitable for human eye interpretation. And can meet the application requirements of low frequency UWB SAR system real-time processing. 3. Based on the visual attention mechanism, the multi-temporal UWB SAR leaf cluster hidden target change detection technology is studied. First of all, from the point of view of visual analysis, the human visual system is simulated, and the Gao Si pyramid model suitable for image analysis is constructed. According to the physiological structure and cognitive characteristics of human visual system, A change detection algorithm for hidden targets in UWB SAR leaf clusters based on visual attention mechanism is proposed, which simplifies complex images to visual focus sets. In order to improve the accuracy of change detection, the visual attention focus is screened layered by using the local neighborhood information of the image and the spatial correlation characteristics of the target. Compared with the widely used change detection technology based on CFAR, the hidden target detection algorithm of UWB SAR leaf cluster based on visual attention mechanism can not only effectively make up for the missed detection problem of CFAR detection method. Moreover, the processing time consumption is only 12% of that of CFAR detection, so it has a great application prospect in practical application.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN958
本文编号：2486311
[Abstract]:Low frequency ultra-broadband synthetic aperture radar (Uitra Wide Band Synthetic Aperture Radar,UWB SAR) has been widely used in civil and military fields because of its strong blade cluster penetration ability. However, with the improvement of the resolution of UWB SAR system and the continuous expansion of its application range, the demand for fast hidden targets and accurate detection of UWB SAR leaf clusters is becoming more and more strong. Therefore, starting with the visual attention mechanism of human visual system, this paper analyzes and establishes a visual attention model suitable for low frequency UWB SAR images, and proposes an unsupervised, fully automatic and efficient method for detecting the change of hidden targets in low frequency UWB SAR leaf clusters. Good detection results have been obtained. The main work and innovations of this paper are as follows: 1. The feasibility analysis of the combination of visual attention mechanism and image processing. Based on the existing data of human visual system, this paper studies the working mechanism of human visual system from the biological point of view, and summarizes the main characteristics of its information processing. This paper discusses the feasibility of applying visual attention mechanism to low frequency UWB SAR image processing, and constructs a visual attention model suitable for low frequency UWB SAR image processing. It provides a theoretical basis for the later research of low frequency UWB SAR change detection technology combined with visual attention mechanism. 2, low frequency UWB SAR image data preprocessing, that is, image visualization technology research. Firstly, combined with the human visual characteristics, the visual basis of low frequency UWB SAR image visualization is determined, and the inherent data statistical characteristics and gray distribution model of low frequency UWB SAR image are analyzed, and the characteristics of low frequency UWB SAR image are analyzed. A fast visualization algorithm for low frequency UWB SAR is proposed, which accords with visual characteristics. Under the same conditions, compared with the traditional logarithmic mapping method and linear mapping method, the results obtained by this method have excellent performance in the assessment of equivalent visual number and processing time, and are more suitable for human eye interpretation. And can meet the application requirements of low frequency UWB SAR system real-time processing. 3. Based on the visual attention mechanism, the multi-temporal UWB SAR leaf cluster hidden target change detection technology is studied. First of all, from the point of view of visual analysis, the human visual system is simulated, and the Gao Si pyramid model suitable for image analysis is constructed. According to the physiological structure and cognitive characteristics of human visual system, A change detection algorithm for hidden targets in UWB SAR leaf clusters based on visual attention mechanism is proposed, which simplifies complex images to visual focus sets. In order to improve the accuracy of change detection, the visual attention focus is screened layered by using the local neighborhood information of the image and the spatial correlation characteristics of the target. Compared with the widely used change detection technology based on CFAR, the hidden target detection algorithm of UWB SAR leaf cluster based on visual attention mechanism can not only effectively make up for the missed detection problem of CFAR detection method. Moreover, the processing time consumption is only 12% of that of CFAR detection, so it has a great application prospect in practical application.
【学位授予单位】：国防科学技术大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TN958
【参考文献】
相关期刊论文 前5条
1 刘娟妮;彭进业;李大湘;王平;;基于谱残差和多分辨率分析的显著目标检测[J];中国图象图形学报;2011年02期
2 王广学;黄晓涛;周智敏;;基于图像分割的VHF SAR叶簇隐蔽目标差值变化检测[J];电子学报;2010年09期
3 周旭;保铮;;SAR目标特性分析技术[J];计算机工程与科学;2008年07期
4 张鹏,王润生;基于视点转移和视区追踪的图像显著区域检测[J];软件学报;2004年06期
5 徐孟侠;图像编码的进展[J];通信学报;1993年02期
相关博士学位论文 前1条
1 钟家强;基于多时相遥感图像的变化检测[D];国防科学技术大学;2005年
，本文编号：2486311
本文链接：https://www.wllwen.com/kejilunwen/wltx/2486311.html
