﻿ 基于改进MTI算法的视频图像空间目标检测
 光电工程  2018, Vol. 45 Issue (8): 180048      DOI: 10.12086/oee.2018.180048

1. 国防科技大学微纳卫星工程中心，湖南 长沙 410073;
2. 武汉大学遥感信息工程学院，湖北 武汉 430079

Space objects detection in video satellite images using improved MTI algorithm
Luo Zhenjie1, Zeng Guoqiang2
1. Micro & Nano Satellite Engineering Center, National University of Defense Technology, Changsha, Hunan 410073, China;
2. School of Remote Sensing and Information Engineering, Wuhan University, Wuhan, Hubei 430079, China
Abstract: An improved MTI algorithm is proposed in this paper to solve the problem of space objects detection in video satellite images. In order to detect the inconsecutive target's trajectory, at the beginning of the algorithm we set a special preprocessing which is called pixel's feeling domain. To reduce the time of the algorithm, we simplified the time projection part of the classic MTI algorithm, which is used to restrain the background. Finally, targets trajectories are obtained through connected domain detection. The experimental results show that, the improved MTI algorithm can effectively eliminate the background and is suitable for the inconsecutive target's trajectory detection. In addition, the algorithm's processing speed almost meets the real-time task.
Keywords: space objects detection    MTI algorithm    pixel's feeling domain    time projection

1 引言

2 基于改进MTI的空间目标检测算法 2.1 算法整体框架

 图 1 SBV经典检测算法流程框架 Fig. 1 The classical detection algorithm framework of SBV

 图 2 本文算法流程框架 Fig. 2 The algorithm framework in this paper

2.2 改进的MTI方法

1) 像素“感受域”预处理

 图 3 相邻两帧的目标位置 Fig. 3 Target's position in adjacent frames

 图 4 相邻两帧的目标位置和“感受域”示意图 Fig. 4 Target's position in adjacent frames and the middle pixel's feeling domain

2) 像素时序信号投影

 ${{\mathit{\boldsymbol{I}}}_{\max }}(x, y) = \max \{ {{\mathit{\boldsymbol{I}}}_k}(x, y)\} ,$ (1)
 ${{\mathit{\boldsymbol{I}}}_{{\rm{mean}}}}(x, y) = \frac{1}{N}\sum\limits_{k = 1}^N {{{\mathit{\boldsymbol{I}}}_k}(x, y)} ,$ (2)
 ${\mathit{\boldsymbol{T}}}(x, y) = \mathop {\arg \max }\limits_k \{ {{\mathit{\boldsymbol{I}}}_k}(x, y)\} ,$ (3)

 ${S_{xx}} = \frac{1}{n}\sum\limits_{i = 1}^n {{{({x_i} - \bar x)}^2}} ,$ (9)
 ${S_{yy}} = \frac{1}{n}\sum\limits_{i = 1}^n {{{({y_i} - \bar y)}^2}} ,$ (10)
 ${S_{xy}} = \frac{1}{N}\sum\limits_{i = 1}^N {({x_i} - \bar x)({y_i} - \bar y)} ,$ (11)
 $L_{\max }^2 = 8({S_{xx}} + {S_{yy}} + C),$ (12)
 $L_{\min }^2 = 8({S_{xx}} + {S_{yy}} - C),$ (13)
 $C = \sqrt {{{({S_{xx}} - {S_{yy}})}^2} + 4S_{xy}^2} ,$ (14)
 $e = \sqrt {1 - \frac{{L_{\min }^2}}{{L_{\max }^2}}} 。$ (15)
2.4 目标条痕端点亚像素定位

 $\left\{ {\begin{array}{*{20}{c}} {{x_k} = \frac{{\sum\limits_{{x_i} \in {\mathit{\Omega } _k}} {{x_i}f({x_i}, {y_i})} }}{{\sum\limits_{{x_i} \in {\mathit{\Omega } _k}} {f({x_i}, {y_i})} }}} \\ {{y_k} = \frac{{\sum\limits_{{y_i} \in {\mathit{\Omega } _k}} {{y_i}f({x_i}, {y_i})} }}{{\sum\limits_{{y_i} \in {\mathit{\Omega } _k}} {f({x_i}, {y_i})} }}} \end{array}} \right.。$ (16)

3 实验分析 3.1 算法检测效果

 图 5 某一帧视频图像截图。(a)视频1；(b)视频2 Fig. 5 A frame of video image capture. (a) Video 1; (b) Video 2

 图 6 本文算法对两段视频的检测结果图。(a)视频1；(b)视频2 Fig. 6 The detection results of the algorithm in this paper. (a) The detection result of video 1; (b) The detection result of video 2
3.2 有无像素“感受域”设置的对比

 图 7 有无设置像素“感受域”的对比。(a)视频1设置像素“感受域”的目标轨迹检测效果局部放大图；(b)视频2设置像素“感受域”的目标轨迹检测效果局部放大图；(c)视频1无设置像素“感受域”的目标轨迹检测效果局部放大图；(d)视频2无设置像素“感受域”的目标检测效果局部放大图 Fig. 7 Contrast before and after setting pixel's feeling domain. (a) Partial map of the object trajectory while using pixel's feeling domain setting in video 1; (b) Partial map of the object trajectory while using pixel's feeling domain setting in video 2; (c) Partial map of the object trajectory while no pixel's feeling domain setting in video 1; (d) Partial map of the object trajectory while using pixel's feeling domain setting in video 2

3.3 算法耗时对比

 帧数 文献[9]对经典MTI的修改方法 本文算法 视频1/s 视频2/s 视频1/s 视频2/s 10 10.67 10.28 0.80 0.73 30 13.51 12.68 1.88 1.85 50 15.74 14.12 3.03 3.00

3.4 算法定位精度评价

 视频序号 人工标注 本文算法 起始帧位置 结束帧位置 起始帧位置 结束帧位置 1 x:455, y:68 x:474, y:96 x:454.13, y:67.74 x:474.03, y:96.04 2 x:51, y:98 x:73, y:155 x:49.90, y:98.07 x:73.68, y:154.98

4 结论

 [1] Scott R L, Wallace B, Bedard D. Space-Based Observations of Satellites from the MOST Microsatellite[R]. Ottawa: Defence R & D Canada, 2006. http: //www. researchgate. net/publication/265232484_Space-Based_Observations_of_Satellites_from_the_MOST_Microsatellite [2] Scott A, Hackett J, Man K. On-orbit results for canada's sapphire optical payload[C]//Advanced Maui Optical and Space Surveillance Technologies, Maui, 2013. http: //www. ncbi. nlm. nih. gov/pubmed/4791780 [3] Wallace B, Scott R, Bedard D, et al. The near-earth orbit surveillance satellite[J]. Proceedings of SPIE, 2006, 6265: 626526. DOI:10.1117/12.672090 [4] Simms L M, Phillion D, De Vries W, et al. Orbit refinement with the STARE telescope[J]. Journal of Small Satellites, 2013, 2(2): 235-251. [5] Anderson J C, Downs G S, Trepagnier P C. Signal processor for space-based visible sensing[J]. Proceedings of SPIE, 1991, 1479: 78-92. [6] Sharma J, von Braun C, Gaposchkin E M. Space-based visible data reduction[J]. Journal of Guidance, Control, and Dynamics, 2000, 23(1): 170-174. [7] Simms L, Riot V, De Vries W, et al. Optical payload for the STARE mission[J]. Proceedings of SPIE, 2011. [8] Shen T S, Guo S J, Xi X L, et al. Timeout controlling for moving object orbit detecting under starry-sky background[J]. Opto-Electronic Engineering, 2016, 43(6): 1-6. 沈同圣, 郭少军, 奚晓梁, 等. 星空运动小目标轨迹检测超时控制[J]. 光电工程, 2016, 43(6): 1-6. DOI:10.3969/j.issn.1003-501X.2016.06.001 [9] Zhang J, Lou S L, Ren J C. Space target detection algorithm based on sequential multi-frames projection[J]. Laser & Infrared, 2015, 45(1): 88-93. 张健, 娄树理, 任建存. 基于时序多帧投影的空间目标检测算法[J]. 激光与红外, 2015, 45(1): 88-93. [10] Liu F, Xi X L, Shen T S. Space small targets detection based on maximum projection and quick registration[J]. Infrared and Laser Engineering, 2016, 45(11): 138-143. 刘峰, 奚晓梁, 沈同圣. 基于最大值投影和快速配准的空间小目标检测[J]. 红外与激光工程, 2016, 45(11): 138-143. [11] Oniga F, Miron M, Danescu R, et al. Automatic recognition of low earth orbit objects from image sequences[C]// Proceedings of 2011 IEEE International Conference on Intelligent Computer Communication and Processing, Cluj-Napoca, 2011: 335-338. http: //ieeexplore. ieee. org/xpls/icp. jsp?arnumber=6047892