Citation: | Yan Lingjie, Huang Yongmei, Zhang Yahui, et al. Research on the application of RANSAC algorithm in electro-optical tracking of space targets[J]. Opto-Electronic Engineering, 2019, 46(11): 180540. doi: 10.12086/oee.2019.180540 |
[1] | 马佳光.捕获跟踪与瞄准系统的基本技术问题[J].光学工程, 1989(3): 1-42. Ma J G. The basic technologies of the acquisition, tracking and pointing systems[J]. Opto-Electronic Engineering, 1989(3): 1-42. |
[2] | 蔡怀宇, 丁蕾, 黄战华, 等.球幕点目标投影跟踪系统的精确标定方法[J].光电工程, 2018, 45(8): 170656. doi: 10.12086/oee.2018.170656 Cai H Y, Ding L, Huang Z H, et al. An accurate calibration method of the ball screen projection point targets tracking system[J]. Opto-Electronic Engineering, 2018, 45(8): 170656. doi: 10.12086/oee.2018.170656 |
[3] | 张沛露, 王建军.车载跟瞄设备低轨卫星预测方法研究[J].数学的实践与认识, 2015, 45(6): 128-132. Zhang P L, Wang J J. Research of LEO satellite orbit prediction for vehicle-borne optical measuring equipment[J]. Mathematics in Practice and Theory, 2015, 45(6): 128-132. |
[4] | 李涛, 张金成.单机动目标跟踪精度分析[J].弹箭与制导学报, 2004, 24(2): 94-96. doi: 10.3969/j.issn.1673-9728.2004.02.031 Li T, Zhang J C. The tracking accurancy analysis of single maneuvering targets[J]. Journal of Projectiles, Rockets, Missiles and Guidance, 2004, 24(2): 94-96. doi: 10.3969/j.issn.1673-9728.2004.02.031 |
[5] | 魏刚, 江传富, 杨坤涛.方位角准确预测法在光电跟踪中的应用研究[J].光电工程, 2006, 33(5): 6-11. doi: 10.3969/j.issn.1003-501X.2006.05.002 Wei G, Jiang C F, Yang K T. Precision azimuth prediction method for electro-optical tracking[J]. Opto-Electronic Engineering, 2006, 33(5): 6-11. doi: 10.3969/j.issn.1003-501X.2006.05.002 |
[6] | 潘晓刚, 周海银, 王炯琦, 等.基于单站短弧段光学观测的低轨卫星轨道预报算法[J].天文学报, 2009, 50(4): 445-458. doi: 10.3321/j.issn:0001-5245.2009.04.011 Pan X G, Zhou H Y, Wang J Q, et al. Orbit prediction algorithm of LEO satellite based on optical measurement in short arc with single station[J]. Acta Astronomica Sinica, 2009, 50(4): 445-458. doi: 10.3321/j.issn:0001-5245.2009.04.011 |
[7] | Choi S, Kim T, Yu W. Robust video stabilization to outlier motion using adaptive RANSAC[C]//2009 IEEE/RSJ International Conference on Intelligent Robots and Systems, St. Louis, MO, USA, 2009: 1897-1902. |
[8] | Chum O, Matas J. Optimal randomized RANSAC[J]. IEEE Transactions on Pattern Analysis and Machine Intelligence, 2008, 30(8): 1472-1482. doi: 10.1109/TPAMI.2007.70787 |
[9] | 黄宗福, 汪金真, 陈曾平.光电探测中空间目标和恒星目标运动特性分析[J].光电工程, 2012, 39(4): 67-72. doi: 10.3969/j.issn.1003-501X.2012.04.012 Huang Z F, Wang J Z, Chen Z P. Motion characteristics analysis of space target and stellar target in opto-electronic observation[J]. Opto-Electronic Engineering, 2012, 39(4): 67-72. doi: 10.3969/j.issn.1003-501X.2012.04.012 |
[10] | 罗浩, 毛银盾, 于涌, 等.利用超大视场光电望远镜观测GEO中的目标识别方法[J].光电工程, 2017, 44(4): 418-426. doi: 10.3969/j.issn.1003-501X.2017.04.006 Luo H, Mao Y D, Yu Y, et al. A method of GEO targets recognition in wide-field opto-electronic telescope observation[J]. Opto-Electronic Engineering, 2017, 44(4): 418-426. doi: 10.3969/j.issn.1003-501X.2017.04.006 |
[11] | 岑明, 傅承毓, 刘兴法, 等.误差空间估计的卫星跟踪位置预测[J].光电工程, 2007, 34(6): 15-19. doi: 10.3969/j.issn.1003-501X.2007.06.004 Cen M, Fu C Y, Liu X F, et al. Position prediction method for satellite tracking based on error-space estimate[J]. Opto-Electronic Engineering, 2007, 34(6): 15-19. doi: 10.3969/j.issn.1003-501X.2007.06.004 |
[12] | 程卫亮, 王向军, 万子敬, 等.压缩域目标跟踪算法在小型化DSP平台上的研究与实现[J].光电工程, 2017, 44(10): 972-982. doi: 10.3969/j.issn.1003-501X.2017.10.005 Cheng W L, Wang X J, Wan Z J, et al. Research and implementation of target tracking algorithm in compression domain on miniaturized DSP platform[J]. Opto-Electronic Engineering, 2017, 44(10): 972-982. doi: 10.3969/j.issn.1003-501X.2017.10.005 |
[13] | 李振伟, 张涛, 孙明国.星空背景下空间目标的快速识别与精密定位[J].光学精密工程, 2015, 23(2): 589-599. Li Z W, Zhang T, Sun M G. Fast recognition and precise orientation of space objects in star background[J]. Optics and Precision Engineering, 2015, 23(2): 589-599. |
[14] | Xu M, Lu J. Distributed RANSAC for the robust estimation of three-dimensional reconstruction[J]. IET Computer Vision, 2012, 6(4): 324-333. doi: 10.1049/iet-cvi.2010.0223 |
[15] | 陈付幸, 王润生.基于预检验的快速随机抽样一致性算法[J].软件学报, 2005, 16(8): 1431-1437. Chen F X, Wang R S. Fast RANSAC with preview model parameters evaluation[J]. Journal of Software, 2005, 16(8): 1431-1437. |
[16] | Hast A, Nysjö J, Marchetti A. Optimal RANSAC-towards a repeatable algorithm for finding the optimal set[J]. Journal of WSCG, 2013, 21(1): 21-30. |
[17] | Chum O, Matas J, Kittler J. Locally optimized RANSAC[C]//Proceedings of 25th DAGM Symposiumon Pattern Recognition, Magdeburg, Germany, 2003, 2781: 236-243. |
[18] | Subbarao R, Meer P. Beyond RANSAC: user independent robust regression[C]//2006 Conference on Computer Vision and Pattern Recognition Workshop (CVPRW'06), New York, NY, USA, 2006. |
[19] | 肖春宝, 冯大政, 冯祥卫.重抽样优化的快速随机抽样一致性算法[J].计算机辅助设计与图形学学报, 2016, 28(4): 606-613. doi: 10.3969/j.issn.1003-9775.2016.04.010 Xiao C B, Feng D Z, Feng X W. Fast RANSAC algorithm with resample optimization[J]. Journal of Computer-Aided Design & Computer Graphics, 2016, 28(4): 606-613. doi: 10.3969/j.issn.1003-9775.2016.04.010 |
Overview: Monitoring and tracking the space debris (referred to as space target) is an important work of electro-optic tracking system. When measuring and tracking space targets in medium and high orbits, due to the long distance between the targets and observation station, the equivalent magnitude is relatively high. In order to enhance the detection capacity of tracking system, the aperture of electro-optic detectors is generally designed to be large. Meanwhile, the field of view of electro-optic detectors is usually small, namely angular component level, for the sake of suppressing the stray light in atmospheric channel. When the space target is shielded by clouds or enters the shadows or penumbra of the earth, the target image cannot be extracted from the field of view of CCD (charge coupled device) camera, the closed-loop tracking of the system cannot be barely work in severe cases. In this case, theoretical orbital data can be used to guide the mount of the system, which keeps track of the target in a short time. However, a large number of high-precision observation data in previous tracking process is discarded. In the absence of multi-station intersection measurement and orbit determination for space targets, guiding and tracking by predicting trajectory is an important way to solve this problem. In this paper, a real-time prediction and tracking algorithm based on short-arc observation data is studied in the background of single station which cannot be intersected.
Random sample consensus (RANSAC) algorithm, which has been widely used in feature extraction in computer vision, is introduced in this paper to achieve higher prediction accuracy. The loss function of RANSAC algorithm is improved according to the distribution of observed data. Considering the fact that smaller errors are usually caused by noise from the interior point and larger errors may be affected by external points, and the boundary between the inner points and the outer points is usually imprecise in practice, the error within the threshold range is taken into account to increase the penalty for larger errors and decrease the penalty for smaller errors. What's more, the improved loss function has a continuous first derivative and varies more gently near the threshold, the sensitivity of the algorithm to threshold is further reduced. The improved algorithm is called WRANSAC algorithm.
The WRANSAC algorithm is proposed according to the distribution of observed data, which is used to deal with the limited observation data in real time to track the space target. After the algorithm is adopted, the fault tolerance of observation data is improved and the sensitivity of the model is reduced. The accuracy and robustness of the prediction results are much better than that of the least squares method. The validity of the WRANSAC algorithm is proved by the comparison between the predicted trajectory and the actual trajectory.
Minimum sampling frequency changing with data error rate
Loss function curve
Flow chart of the WRANSAC algorithm
Histogram of interior point cost function
Histogram of internal point weights
Prediction of 72 minutes
The error distribution for the prediction of 72 minutes