大视距非合作目标视觉位姿测量系统关键技术研究

发布时间：2019-01-17 09:21

【摘要】：在航天器在轨服务、攻防对抗、以及空间垃圾处理等任务中需要测量两个物体之间的相对位姿。由于测量距离由远及近跨度比较大,测量目标上不能安装对应的合作靶标,因此需要进行大视距非合作目标位姿测量。基于视觉的位姿测量系统因为具有系统体积小,功耗小以及同时可以获得实时图像等优点而受到广泛关注。但是由于非合作目标先验知识较少,大视距下视觉位姿测量精度变低等导致其不能满足实际要求。针对这些缺点,即如何在被测目标已知先验知识较少情况下能够实现位姿测量,以及在大视距下提高位姿测量精度成为研究难点。针对这些难点本文主要从以下几个方面进行了研究,具体内容如下:针对非合作目标先验知识较少情况下能够实现位姿测量的问题,本文提出基于垂直直线特征的双目视觉位姿测量算法。该算法仅需已知被测目标上两条直线相互垂直关系,利用四元素法和直线的垂直关系求解出被测目标的相对位姿。仿真实验表明在0.1和0.5像素标准差的图像噪声影响下,位姿测量算法在0.3m~25m具有较好的测量精度。为了使得位姿测量精度能够满足大视距的要求,需要高精度地标定出位姿测量模型中的各种参数。针对大视距中广角镜头的畸变较大的问题,本文采用了消隐点共线约束优化算法来提高交比不变性逐点畸变校正精度。该算法首先利用畸变较小的中心区域图像点根据交比不变性来逐点计算周围图像点的理想值,然后利用消隐点共线约束算法来优化计算图像点,最终提高了畸变校正精度。摄像机内部参数和外部参数误差耦合使得摄像机相对于标定靶标不同姿态、不同位置时标定精度不同。为了提高摄像机内外参数标定精度,需要减小摄像机内外参数标定误差之间的耦合关系对标定参数精度的影响。本文提出了内外参数解耦算法,该算法根据内外参数误差存在的线性耦合关系,利用靶标平移后的多次标定结果,对摄像机像面主点、摄像机焦距和平移向量的标定误差进行修正。实验结果表明,在噪声和畸变影响下,解耦算法减小了内外参数误差耦合对标定参数精度的影响,提高了参数标定精度。在航天某实验卫星预研的硬件平台上验证了本文所提出的算法。该平台包括两个摄像机、被测目标、嵌入式图像处理硬件和地面验证设备。嵌入式硬件中写入本文所提出的位姿测量算法。地面验证设备采用光学暗室来模拟太空环境,采用六自由度运动系统来检测视觉系统测量精度。最终实验数据表明非合作目标视觉位姿测量精度满足指标要求,从而也验证了本文所提算法的有效性。
[Abstract]:It is necessary to measure the relative position and orientation of two objects in the missions of spacecraft in orbit service, attack and defense countermeasures, and space garbage disposal. Because the measurement distance is large from far to short span, the corresponding cooperative target can not be installed on the measurement target, so it is necessary to measure the position and attitude of the non-cooperative target with large visual range. The position and pose measurement system based on vision has attracted much attention because of its advantages of small size, low power consumption and the ability to obtain real-time images at the same time. However, due to the lack of prior knowledge of non-cooperative targets and the low accuracy of vision pose measurement under large visual range, it can not meet the practical requirements. In view of these shortcomings, how to realize the pose and attitude measurement under the condition of the known prior knowledge of the target under test, and how to improve the precision of the position and attitude measurement under the large visual range become the research difficulties. In view of these difficulties, this paper mainly studies the following aspects, the specific contents are as follows: aiming at the problem that the position and attitude measurement can be realized under the condition of little prior knowledge of non-cooperative target, In this paper, a binocular vision pose measurement algorithm based on vertical line features is proposed. The algorithm only needs to know the vertical relationship between two lines on the target under test, and solves the relative pose of the target by using the method of four elements and the vertical relation of the line. The simulation results show that the pose measurement algorithm has a good accuracy in 0.3m~25m under the influence of image noise of 0.1 and 0.5 pixel standard deviation. In order to make the precision of pose measurement meet the requirement of large visual range, it is necessary to calibrate the parameters in the model of pose measurement with high precision. Aiming at the problem of large distortion of wide-angle lens in large visual range, the collinear constrained optimization algorithm of blanking point is adopted to improve the accuracy of point by point distortion correction of cross-ratio invariance. The algorithm firstly calculates the ideal value of the surrounding image point by using the image points of the center region with small distortion according to the invariance of the cross ratio, and then optimizes the calculation of the image points by using the collinear constraint algorithm of the blanking points to improve the accuracy of distortion correction. The coupling of the internal and external parameters of the camera makes the camera with different attitude to the target, and the calibration accuracy is different in different positions. In order to improve the accuracy of camera calibration, it is necessary to reduce the effect of coupling relationship between camera internal and external parameters calibration error on the calibration accuracy. In this paper, a decoupling algorithm for internal and external parameters is proposed. According to the linear coupling relationship between the internal and external parameter errors, the camera image plane is determined by the multiple calibration results of the target translation. The calibration error of camera focal length and translation vector is corrected. The experimental results show that under the influence of noise and distortion, the decoupling algorithm reduces the effect of the error coupling of internal and external parameters on the calibration accuracy and improves the calibration accuracy. The algorithm proposed in this paper is verified on the hardware platform of an experimental satellite. The platform includes two cameras, target to be tested, embedded image processing hardware and ground verification equipment. The position and pose measurement algorithm proposed in this paper is written in embedded hardware. The optical darkroom is used to simulate the space environment and the 6-DOF motion system is used to detect the measuring accuracy of the visual system. Finally, the experimental data show that the accuracy of the non-cooperative target vision pose measurement meets the requirements of the target, which also verifies the effectiveness of the proposed algorithm.
【学位授予单位】：哈尔滨工业大学
【学位级别】：博士
【学位授予年份】：2016
【分类号】：TP391.41

【相似文献】