【作者】 王征；

【导师】 汪洪源；

【作者基本信息】 哈尔滨工业大学 ， 光学工程， 2016， 硕士

【摘要】 随着各国空间技术的迅猛发展,发射进入太空的空间目标数量越来越多,为有效提升空间态势感知能力、迫切需要基于获取的图像数据开展空间目标的跟踪性能研究。在空间目标跟踪过程中,由于成像链路各环节误差、星载平台抖动等原因造成图像质量退化,进而影响目标的跟踪精度。为精确分析图像质量对跟踪精度的影响关系,构建了科学的图像质量评价体系,提出了可行的空间目标跟踪精度分析方法,并通过数值优化算法揭示了图像质量对跟踪精度的影响规律。具体研究内容包括以下几方面:首先,开展了序列图像质量评价研究。基于图像特征分析,构建了图像质量评价的指标体系,具体评价指标包括易于提取及定量度量的灰度、纹理、能量等18种图像质量评价参数。同时,针对图像质量参数间相关性突出的问题,基于主成份分析方法完成了参数的合理优选,选取了累积贡献率超过90%的3个主成份,为后续研究图像质量对目标跟踪性能的影响关系提供输入参数。其次,进行了目标跟踪精度的仿真分析。根据序列图像目标跟踪任务需求,提出基于图像特征提取与匹配的目标跟踪精度分析方法,经过对图像源的特性分析,选取较成熟的SIFT算法及Harris角点检测算法进行图像匹配,通过序列图像目标跟踪精度的仿真分析,为后续研究图像质量对目标跟踪性能的影响分析提供输出参数。最后,开展了序列图像质量对跟踪精度影响分析。结合输入与输出参数的特点,将图像质量对跟踪精度的影响关系研究看做黑箱问题,采用BP神经网络算法从整体上建立了系统输入对输出影响关系模型,并以仿真生成的空间目标图像为例,计算获取了图像质量对跟踪精度的绝对影响系数。总之,通过序列图像质量对目标跟踪精度的影响研究,建立了图像质量评价模型、提出了目标跟踪精度分析方法,构建了序列图像质量对目标跟踪精度的绝对影响关系,研究结果为成像探测系统的方案设计与优化、在轨图像质量提升与评价提供重要的理论依据及技术手段。更多还原

【Abstract】 Along with the improvement of space technology, the quantity of object launched into space has become biger and biger. In order to improve the ability of space, we need the research on space object tracking that is based on gained data of image. On the process of object observation by satellite, there will be degeneration of image quality and error of object tracking because of error in imaging links and shaking of satellite-borne platform. In order to analyze the influence of image quality on tracking accuracy, scientific evaluation system for image quality is constructed and proposal analysis method about tracking accuracy is presented. And influence law about image quality and tracking accuracy is gained by way of numerical optimization. Specific research content include following aspects.First of all, the evaluation of sequence image quality is carried out. Construct index system of the image quality assessment and find 18 evaluation parameters that are easily extracted and quantitatively measured, such as grayscale, texture and energy. Aiming at the problems of correlation among parameters, the optimization of them is finished by principal component analysis. The three principal components whose accumulative contribution rate surpasses ninety percent are chosen as input parameters for next research on the influence of the object tracking accuracy of sequence image.Secondly, simulation analysis on object tracking accuracy is presented. According to the requirement of sequence image tracking, the analysis method for object tracking accuracy is presented based on the extracting and matching of image character. Through the character analysis of image source, image is matched by mature algorithms such as SIFT algorithm and Harris corner detection algorithm. Comparing the tracking accuracies of two algorithms, scientific evaluation for object tracking algorithms is presented. At the same time, the tracking accuracies are used as output parameters for next research on the influence of the object tracking accuracy of image quality.Finally, the analysis of tracking accuracy of image quality is carried out. Combing the characteristics of input and output parameters, the research about influence of image quality on the object tracking accuracy is regarded as black-box problem. BP neural network algorithm is adopted to construct the model of relationship between the input and output, and use the simulation generated image as example to gain the absolute influence coefficient about image quality and tracking accuracy.All in all, by research on influence of image quality on the object tracking accuracy, the evaluation model of image quality is constructed, analysis methods about object tracking accuracy is presented and absolute influence coffiicent about sequence image quality and object tracking accuracy is presented in this paper. The result of paper supply important theory basis and technical means for design and optimization in imaging detection system and for promotion and evaluation in image quality.更多还原