【作者】 刘和祥；

【导师】 边信黔；

【作者基本信息】 哈尔滨工程大学 ， 控制理论与控制工程， 2009， 博士

【摘要】 自主水下航行器(简称:AUV)是高技术的集成体,AUV的水下回收问题是其诸多关键技术中迫切需要研究、解决的关键技术之一。为此,本文深入丌展了AUV回收过程中的视觉测量定位以及AUV与水下回收平台对接的关键性技术研究。1、针对水下机器视觉实现困难、图像处理十分复杂的特点,本文将高性能的数字摄像机与先进的HALCON图像处理软件有机结合,在程序语言接口、图像采集接口、数据格式和摄像机触发控制上进行软硬件匹配,使之能够正确、实时地获取水下空间的视觉信息;针对AUV运动时双摄像机曝光不同步所产生的异位采集问题,设计了摄像机外触发同步器,利用此装置对双目视觉摄像机成功进行了外触发实验;针对水下照明功率问题和光源发散问题,设计了由发光二极管组合而成的多种易于辨别和区分的形状阵列,解决了水下对接的目标显示问题。2、针对回收过程中视觉定位的误差问题,采用一种基于投影矩阵的摄像机水下立体标定方法,提高了目标定位过程中的测量精度。该方法通过线性模型分析计算得出摄像机的内外参数,用最大似然法对其进行非线性优化并考虑镜头畸变的目标函数,最后求出每个摄像机的内、外参数,标定方法准确性高,不需昂贵的器材,简单实用,比较适合水下现场标定;针对标定板图像中的特征圆分割与提取问题,采用最优阈值分割算法分割标定板图像,提出基于区域标记的噪声点消除算法,消除水下噪声干扰,精确提取出所有特征圆。为检验标定结果的准确性,进行水下目标靶三维测量实验,实验测量的绝对误差在±5mm毫米之内。3、针对水下图像的噪声问题和细节模糊问题,设计了高斯平滑滤波器对水下图像进行平滑滤波处理,采用形态学滤波消除物体边界点,利用灰度线性变换的方法,增加水下图像的对比度,使图像细节更容易看清;针对水下目标的轮廓提取问题,采用的Canny亚像素边缘检测算法提取水下图像中所有物体的边缘,利用改进的Snake模型算法在所有物体的边缘中精确提取封闭的目标图形轮廓。4、针对水下回收平台的目标识别问题,设计了多个内部形状不同的圆形目标安放在回收舱底部,采用形状模型匹配的方法来寻找所有的圆形目标;针对目标图像的模型搜索问题,提出了基于规则的模型搜索策略,通过调整和优化搜索参数,提高了模型搜索的速度和准确性;针对水下回收平台的圆形目标定位问题,通过计算每一个目标点和镜头焦点连线与成像平面的交点坐标,结合标定后摄像机的内参和外参,进行视差计算和坐标转化,精确计算出所有目标点与摄像机的相对坐标。为验证视觉定位算法的可行性,进行了夜间水下实验,成功模拟AUV在深水环境中对目标的精确定位。5、针对AUV运动过程中存在的严重非线性和不确定性,在建立AUV六自由度数学模型的基础上,设计了用于AUV回收的H_∞鲁棒控制器,并对AUV的运动控制进行了仿真。为了验证视觉导引算法的正确性,根据设定的比例,结合11自由度实验台架,搭建了AUV回收物理模拟实验平台,设计和开发了视觉算法程序和AUV回收控制程序。在模拟实验中,将AUV水下对接分为四自由度接近、悬停定位、逼近以及入舱操作三个阶段。在夜暗环境下成功进行了基于视觉的回收模拟验证,为下一阶段的AUV回收水池试验打下了基础。本文为了实现AUV水下自主回收,对视觉导引方法进行了探索性地研究,实现了定位偏差达到毫米级的摄像机水下立体标定,在立体视觉的匹配过程中,提出了基于规则的形状模型搜索策略,利用视差法对目标点进行定位计算。通过缩尺度模拟实验,验证了本论文视觉导引方法的有效性和适用性。本论文的研究成果和结论对于AUV安全、可靠、高效地完成远程航海与地形勘察作业使命,具有重要的理论意义和实际应用价值。更多还原

【Abstract】 Autonomous Underwater Vehicle(AUV) is an important research field which is integrated with many an oceanic high technology.The reclaim problem is one of the most demanding technologies of AUV that need to be researched and tackled.In this paper,the technologies of vision guidance of AUV docking are researched primarily.Firstly,the digital camera with high performance and advanced HALCON software are combined to overcome the difficulty of underwater machine vision and the complicity of image processing.To obtain the underwater space vision information properly and promptly,the programming language,images collecting, data format and camera triggering controlling are matched through soft and hard ware.The outer triggering synchronization of camera is designed to address the problem of asynchronism collecting which is caused by exposure asynchronism of double-camera when AUV is moving,which can be used to operate outer triggering experiment.The figuring array consisted by luminous diode is designed to deal with the problems of illumination power and light-house radiation,which solved target displaying for AUV docking.Secondly,the method of underwater stereo calibration based on projection homography is designed to resolve error precision of vision position in the process of docking,improve the measurement precision during the process of target position.The method utilize linear model to calculate the optimization of the parameters of camera,and figure the non-linear precision by considering the target function of lens aberration,together with the result of internal and external parameters,which is highly precision,lower cost,simply operated and adapted to underwater local calibration.The algorithm of noising point eliminating based on region marked is proposed by adopting calibration board images through optimized algorithm of threshold segmentation,to settle the problem of feature circle segmentation and abstraction,which can also eliminate the disturb of underwater noise and abstract all feature circles.To verify the veracity of calibration result,the 3-D measuring experiment of underwater target is carried out,the absolute error of the precision degree is±5mm level.Thirdly,due to the problems of noise and detailed fuzzy for underwater images,gauss smoothly filter is developed.The object edge point is eliminated through morphologic filter,and the contrast of underwater images is enhanced and the detail of images is distinguished more easily by using prey linear transform method.The algorithm of Canny sub-pixel edge detecting is introduced to do with the problem of contour abstraction for underwater target,and use the improved snake module to abstract the graphics contour from all the objects precisely.Fourthly,a few circle targets which is different in internal shape is proposed to install on the bottom of the docking cabin to overcome the target recognizing problem during underwater docking process,all circle targets are searched through shape template matching method.Because of model searching problem, the searching tactic is put forward based on regular model,and through adjusting and optimizing parameters,the searching rate and accuracy is improved. According to the target position demanding of the underwater docking platform, the relative coordinate between all the target points and camera is figured out precisely,through calculating the intersection points coordinate of each target and lens and combing the internal and external parameters after calibrating.To validate the practicality of vision position algorithm,the night experiment is made out,and simulates the correct target position for AUV in the underwater environment successfully.Lastly,according to the nonlinear and uncertainty during AUV movement,the H_∞controller is developed for AUV reclaiming,which is based on 6 DOFs of AUV mathematics modeling.To verify the effectiveness of vision guidance algorithm,according to the designed proportion and combing 11 freedom degree of the test shelf,the AUV docking simulating platform is set up,and vision algorithm and docking controlling programming are designed and developed. During the experiment,the docking process is divided into 4 freedom degree approaching phase,suspending position phase and entrance cabin phase.The docking test in the night environment is processed successfully,which can make a good basis for the pool experiment for the AUV docking.In order to achieve autonomous reclaiming technology for AUV,vision guidance method is researched,accomplish camera stereo calibration,and the poison precision degree is mm level.In the process of stereo vision matching,the searching tactic is put forward based on regular model,and calculate the position through vision difference.The effectiveness and practicality of the presented methods are verified through deflating scaling experiment.The research and conclusion of the thesis have a great both theoretical and practical value for the AUV,such as safety,reliability,remotely sailing and landform reconnaissance.更多还原