【作者】 孙剑明；

【导师】 周卫东；

【作者基本信息】 哈尔滨工程大学 ， 导航、制导与控制， 2013， 博士

【摘要】 在舰船上使用天文导航进行比较准确的定位是一个复杂的工作流程，为了保证舰船通过天文导航星图识别方法获取到准确的方位信息，论文研究了在舰船上使用天文导航方式的关键技术方法：包括Delaunay三角剖分的弧度值建立星库，基于顶点剖分弧度集合模糊匹配方法的星图识别，基于层次反应型Agent对海天图像进行云影云迹的去除以及基于改进Hough变换的海天线快速检测算法，实验结果与分析证实了提出的方法能够充分保证舰船在海上利用天文导航信息进行准确的定位，进而通过方位信息矫正惯性导航误差。论文主要工作如下：研究了各种坐标系统下的位置转换关系，讨论了时间系统以及不同时间系统下换算关系，深入的阐述了在已知的标准星库基础上建立星点之间的剖分弧度特征集合。通过筛选SAO星表中星等大于7.0的恒星，建立Delaunay三角剖分特征识别星库。提出了星图识别方法及星图识别准确概率信度方法。首先对获取到的星图进行必要的去噪和复原处理，使用一种全新的模糊边际划分方式将星点逐一检测提取出来，将每个星点进行Delaunay三角剖分，使用基于顶点剖分弧度集合模糊匹配方法的星图识别方法找到与星库中特征集合最相近的那一对恒星，最后提出了一个检验识别正确概率的星图识别信度的分析方法，为判断已经识别出星点的识别正确性。舰船姿态的判断是位置确定的关键，主要利用位于舰首和右舷的红外取景装置所拍摄的海天线倾斜角度来判断船体的横摇角与航向角。首先使用层次反应型Agent高效准确地去除海天线图像中存在的云影云迹干扰，然后使用基于改进Hough变换的海天线快速检测算法快速准确的计算出船体的倾斜值。提出的天文导航系统与惯性导航系统进行误差漂移的比较。首先对惯性导航系统误差产生进行了分析与仿真。然后着重讨论了本天文导航方法存在大气折射误差、图像像素误差等的误差因素。给出了本天文导航方法产生误差的大致范围，该范围小于惯性导航系统在使用一段时间后产生的误差范围。最后给出了通过天顶处赤经与赤纬以及该时刻世界时计算经度与纬度的公式。通过系统性仿真实验分析，结果表明：采用弧度集合模糊识别的方法星识别成功率可达到98%，船舶倾斜角的判断也是正确有效的，基于这些技术手段在舰船上使用天文导航方法来定位定向是可靠有效的。更多还原

【Abstract】 It is very complicated for naval ships to fix positions with astronavigation. Toensure naval ships can acquire accurate position information by astronavigation starpattern recognition， the author in this paper researches some key technologicalmethods in naval ships by astronavigation， which includes the building of stardatabase by radian value of Delaunay triangulation， star recognition based onapproximate matching of vertex subdivision radian set， elimination of clouds in seaand sky images on the basis of reactive Agent of levels and fast detection algorithm ofline between sea and sky according to improved Hough transform. Results ofexperiments and analysis show that methods mentioned above can be used to fixpositions precisely by astronavigation information on naval ships，and then to correctchronic navigation errors by position information. This dissertation will do thefollowings：Present paper research position switch relation in all kinds of coordinatedsystems， and time system and conversion relation in different time systems. And wereveal elaborately characteristic set of subdivision radian between stars is set up inlight of the known standard star database. By screening out fixed stars over7.0fromSAO Star Catalog， we build characteristic recognition star database of Delaunaytriangulation.Star recognition method and exact probability reliability method of it areproposed. First， we need to denoise and recover star image that we get， and extractevery star point by a new method of approximate marginal division， and then doDelaunay triangulation to every star point. We are based on star recognition method ofapproximate matching approach of of vertex subdivision radian set to find that pair offixed stars most similar to characteristic set in star database. Finally， we put forwardan analysis method of star recognition reliability that can test and recognize exactprobability， judge recognition correctness of the star points which have beenrecognized.The attitude judgement of naval ships is vital to fix position. Rake angle of linesbetween sea and sky shot by infrared viewfinder on the bow and starboard is used tojudge rolling angle and course angle of the ship. Reactive Agent of levels is firstapplied to remove clouds in image of sea and sky efficiently and precisely. And thenwe， according to improved Hough transform， use fast detection algorithm of linebetween sea and sky to calculate lean value of the ship.Present paper compare drift errors between Celestial Navigation proposed in thisdissertation and chronic navigation system. First，we analyze and simulate errors thatchronic navigation system makes. Secondly， we exert importance on researching theerror factors， such as atmospheric refraction error， image pixel error， which thisCelestial Navigation approach makes. Thirdly， we reveal the error range made bythis astronavigation approach，and this range is smaller than the one made by chronicnavigation system. Finally， we demonstrate formulations of calculating longitudeand altitude by right ascension and declination of zenith.Present paper use that set of methods proposed in this dissertation to simulate system. There are a sum of recognition data and calculation process when we acquireand recognize a new star image and judge recognition reliability of star image.Combining attitude of the ship， we can get real position of zenith in the star image.Finally， we can calculate longitude and altitude of the ship in the light of world time.By analyzing systematic simulation experiment， we can find that the method ofusing arc set fuzzy identification is very successful because its rate of success canreach98%. At the same time， judgement of rake angle of ships is correct and valid.Therefore， the method of using Celestial Navigation to locate places and directionson ships is reliable and valid.更多还原