【作者】 张晓玲；

【导师】 林玉池；

【作者基本信息】 天津大学 ， 测试计量技术及仪器， 2005， 博士

【摘要】 本研究为本实验室与二炮合作项目“导弹落地阶段动态参数测量技术的研究”的预研。本课题采用双目视觉立体成像技术,建立了适用于导弹航天测量的光电测量系统。在该测量系统中,摄像机以固定方式(非跟踪方式)对着预定空域,将光电经纬仪和弹道相机作用原理部分应用于解析摄影中,实现远距离运动物体的三维轮廓、运行姿态和运动速度的测量。该系统具有对硬件要求低,实现简单,成本低,精度较高,稳定性较高,运用范围广的特点。本论文主要完成以下研究工作:1.详细论述了简化的双目立体视觉测量系统,分别建立了光轴平行测量系统和光轴垂直测量系统的数学模型,并对高速摄像机参数选择进行了理论推导和选择。从光电经纬仪的测量原理及一般双目立体视觉测量系统分析出发,提出了简化的双目视觉系统,并分别建立了光轴平行测量系统和光轴垂直测量系统的数学模型。分析讨论了摄像机摆放位置与摄像机内部参数之间的关系。理论推导了测量用摄像机的内部参数的选择。通过实验,给出了当落弹范围要求不同时,摄像机摆放位置与摄像机内部参数之间的关系,给出了可选用的摄像机参数的部分数据。2.详细论述了两摄像机光轴平行和两摄像机光轴垂直时的摄像机标定方法。从理想透视原理出发,推导并建立了不考虑畸变的理想摄像机成像模型。在此基础上,论述了考虑径向和切向畸变的摄像机模型,并用罚函数法对畸变模型中的摄像机内外部参数进行了分析求解。详细研究了两摄像机光轴平行和光轴垂直的实现条件,并建立了相应的数学模型。通过实验,给出了左右CCD的内部和外部参数以及径向和切向畸变系数等标定结果,并分析了标定的误差,给出了两摄像机光轴平行和光轴垂直的标定图。结果表明,这种标定方法精度很高,X方向的标定误差最大只有±0.05pixel,Y方向的标定误差最大只有±0.07pixel。3.系统研究并分析了对远距离运动物体的图像进行边缘特征提取和立体匹配的方法,并给出了相应的解决方案。在研究三种常用的特征提取方法的基础上,提出以下特征提取方法:用灰度动态范围压缩方法对图像进行灰度预处理,用掩膜的方法将图像从背景中提取出来,用Sobel算子提取出图像的边缘信息。从摄像机的成像平面几何原理出发,用水平扫描线搜索的方法,得到匹配右像点的取值范围;用灰度相关与模板匹配相结合的方法,在取值范围内得到待匹配的右像点坐标;采用相关窗口方法对非边缘像点进行立体匹配,实现了远距离运动物体三维信息的恢复。应用水平顺序匹配约束策略对视差图进行误差检测与校正。通过实验,给出了不同被测物的边缘信息、视差图、经过校正的视差图以及恢复出的三维轮廓。4.详细研究讨论了在两种简化的双目测量系统中炮弹运行姿态和运动速度的求取方法。当采用两光轴平行系统时,利用运动目标表面点到弹轴的距离相等的原理,用最小二乘拟合法得到弹轴矢量。当采用两光轴垂直系统时,应用两种方法来求取弹轴矢量:一种是利用中轴线法推导出炮弹在两个CCD像面上所成的直线方程,通过透射原理得到弹轴矢量;另一种是利用光轴垂直原理求取弹轴矢量。从弹轴矢量出发,推导了姿态角的数学公式。应用三维插值方法,对立体匹配后得到的N个时刻坐标点进行插值,求取运行距离、平均速度。进行了求取炮弹模型在运动过程中的运行轨迹、运行姿态角、运动距离以及运动速度等运动参数的实验。实验结果表明,该测量方法具有较高的精度,能满足远距离运动物体运动参数的测量要求。更多还原

【Abstract】 The research work of the dissertation is the pre-research item of the project“The theoretical and experimental research on dynamic parameter test technology of missile falling to the ground”, cooperated with PLA Second Artillery Forces. An electro-optical measurement system has been developed. Using the imaging technology of the binocular vision, the proposed electro-optical measurement system can measure the attitude and velocity of remote motion object easily. In the measurement system two CCDs are fixed toward the space which the remote moving object will go through, and the working principles of the photoelectric theodolite and the ballistic camera are used in the analytic photogrammetry. It is of the advantages such as low requirement to hardware, easy to implementation, low cost, high accuracy, and extensive applied fields. The research work of the dissertation mainly includes four aspects as follows:1. The simplified binocular stereo vision measurement systems are studied. On the basis of the analysis on general binocular stereo vision measurement system, two kinds of novel simplified binocular stereo vision measurement systems, namely one system in which two cameras’(CCD ) ray axes are parallel and the other in which two cameras’(CCD) ray axes are vertical, is proposed and the corresponding mathematic model is established. The relation of the camera location and its inner parameters is researched and discussed. Based on the above relation, how to select the parameters of the high speed camera employed in the measurement system is analyzed. Through the experiment, the relation of the camera location and its inner parameters is given and some available camera parameters are obtained when the moving object fall to different site on the ground.2. The CCD calibration method of the two simplified binocular stereo vision measurement systems is researched. Starting with perspective theory, the camera imaging model without thinking of distortion is derived and established. The camera imaging model with radial and tangential distortion is given, and the outer and inner parameters of CCD are obtained with the penalty function algorithms. The parallel condition and vertical condition of CCDs’ray axes is studied and performed separately. Experiments give the calibration results such as inner and outer parameter, radial and tangentialdistortion coefficients of the CCD, and calibration figures. The calibration error is also analyzed. The results show that the calibration method is of high precision: the error of X direction is no more than±0.05pixel, and the error of Y direction is no more than±0.07pixel.3. The methods of edge feature extracting and stereo matching on images of the remote moving object are researched and the corresponding settle schemes are given. On the basis of studying three general feature extracting methods, we realize the edge feature extracting as follows: Firstly, the images collected are pre-processed with grey compress method; Secondly, the images are extracted with the mask method; Lastly, the edge information of the images are obtained with the Sobel operator. The stereo matching on images are proceeded as follows: Firstly, The span of right points matching with left points is obtained with the search method of level line; Secondly, the grey correlation method and model matching model are employed to obtain the coordinates of right points from the span; Lastly, the related window method is used to realize the stereo matching of the points which is not edge points, and then the 3-D profile of the moving object can be recovered. With the level sequence matching strategy, the error of disparity image is detected and corrected. A lot of experiments have been done and the results such as the edge information, disparity image, corrected disparity image and recovered 3-D profile of different measured object at different angle, are obtained.4. The implement methods of the attitude and velocity of missile in the two simplified binocular vision measurement system are discussed. When the measurement system in which two CCDs’ray axes are parallel, is employed, starting with the principle of equal interval between the measured object’s surface points and its axis, the axis vector of the object is obtained with the least-square fitting method. When the measurement system in which two CCDs’ray axes are vertical, is adopted, there are two methods to get the axis vector of measured object. One method is to derive the linear equation of the object axis’projection line in CCD image plane with the axle wire method, and then to get the axis vector of the object using the transmission principle. The other method is to adopt the vertical principle of two CCDs’optical axes to obtain the axis vector of measured object. Starting with the axis vector, the attitude angle is derived. The 3-D interpolation method is improved to interpolate the 3-D coordinates of measured model in different times which are obtained by stereo matching. The moving distance of object is gotten by the3-D interpolation, and then the mean velocity is obtained. The experiment to get the motion parameters of missile model such as moving trace, attitude, interval and velocity has been done. The experiment results show that the measurement methods are of high accuracy, and can satisfy the measurement require of remote moving object.更多还原