微型航天探测系统及其目标识别算法研究

【作者】 刘辉；

【导师】 胡炳樑；

【作者基本信息】 中国科学院研究生院（西安光学精密机械研究所） ， 通信与信息系统， 2007， 硕士

【摘要】 随着空间科学技术的发展，航天探测器受到越来越多国家的关注。而体积小、功耗低、成本低的微型航天探测器是目前航天器发展的主要趋势。微型航天器因其目标小、机动性强、可进行编队组网等特点，在军民两用高新技术中受到全世界的广泛关注。在军事领域，可利用微型航天器来完成多项任务，如空间目标成像、空间电子侦察、空间电子干扰和目标拦截等。在非军事领域，微型航天器的一个重要用途是进行空间技术在轨试验。低投入、短周期的微型卫星将给高新技术领域的太空试验带来希望，而高技术演示验证是将研究成果工程化的重要手段，可以加速研究成果向工程研制和应用的转化，降低研制风险，提高使用效益，从而加速技术更新，推动空间技术的发展。由于微小卫星领域的技术发展和更新非常迅速，因此加强高技术演示验证就成为一项极为重要的工作。微型航天探测器主要是对太空中的目标物进行探测成像并主识别，然而由于微型航天探测器所处的环境的背景比较复杂，再加上复杂背景、目标物以及探测器三者之间有相对的运动使得对目标物的识别比较困难。而且微型航天探测器体积小，硬件资源、处理速度和能量都及其有限。论文从微型航天探测器的实际情况出发，在对现有识别方法进行分析比较后，提出了两类新的识别算法：一类是多帧图像识别算法，主要有基于图像配准的边缘提取差分法和基于图像配准的双差分法。它们主要是针对背景、目标物和探测器之间有相对运动提出来的新的算法。这两种算法的关键点在图像配准上。另一类是单帧图像识别算法即改进的连通域法的目标识别算法，该算法对单帧图像进行处理避免了多帧之间由于背景、目标物和探测器之间有相对运动带来的影响，而且需要的硬件资源和计算量都比较小。它首先对探测器采集的图像进行二值化，然后对整幅二值化的图像进行连通域标记，最后用领域法搜索图像中最大的连通域，即目标物，计算出目标物的形心以及目标物方位角、俯仰角和波门的大小等参数。实验表明该算法在处理速度和硬件资源要求上都优于前者以及一些现有的识别算法。项目中，在PC机上用VC实现了探测器与PC机的连接以及开发出满足项目要求的并带有不同目标识别算法的软件系统。针对微型航天探测器的实际特点，用DSP硬件实现改进连通域的目标识别算法，并设计了FPGA实验板模拟CMOS数据源，实现并达到了实时目标识别的要求，获得了较好的识别效果。本文所做的工作主要是完成了微型航天探测器对空间目标物识别预研，还对下一步工作提出了有益的建议。更多还原

【Abstract】 Space detecting is paid attention to by more and more country along with thedeveloping of space technology. Small size, low power and low cost MicrospacecraftTarget Detector(MTD) is the mainly developing detection in the space detecting field.The MTD get worldwide attention in the military affairs and civil hightecnology for itscharacters of small,easily move,can famate and so on. In the military affairs, the MTDcan do many kinds of tasks. For example the imaging of space target, space electronicreconnoiters, electronic interferes and target hold off. In the civil affairs the mostimportant use of MTD is the spacial experiment on the orbit. Small satellit with lowcost and short period will bring hope for spacial experiment in high and newtechnology field, on the other hand the high technology demonstrating and validating isan important way to make the researching production into project. It can accelerate thetransform for research productions to project develop and application, reduce theventure of development, and improve using benefit. So the MTD accelerates the speedof validating the new technology and the speed of technical update and put the spacialtechnology forward. Because the technology in small satellite field developing andrefreshing is very fast, it is a very important work to strengthen high technologydemonstration and validation.The MTD mainly function is imaging and recongnizing the spacial target. In spacebackground, there is relative movement among the target, and MTD. So it is difficult torecognize the target. Further more the MTD’s cubage, hardware resource, processingspeed and power are very limited.Considerating the source of MTD, two kinds of recognising algorithm was shownin the thesis: One is multi-frame image recognising algorithm. It mainly containsimage edge extract and deference base on image registration and double differencebased on image registration. Aimed at resolving the relative movement among thetarget, space background and MTD, the new algorithm was put forward. The core ofthe two algorithm is image registration. Another algorithm is single-frame imagerecognising algorithm, called improved connected recognising algorithm. Thealgorithm only processes one frame image and avoids the influence of the relativemovement among the target,space background and MTD. By the way the hardwareresource and calculation are small. The first processing of algorithm is binary theimage frame MTD, the second step is connecting component labeling the binaryimage, at last one is searching the image with egiht-coneceted way to find the biggestconnected in the image-the target, and caculate out the parameters of azimuth angle,pitching angle and the dimension of the gate. The result of experiment indicates thatthe improved connected algorithm excelled the other algorithms on the processingspeed and the hardware resource requirement. In the project, the connection betweendectecting sensor and computer was realized, and exploit out a software system whichcontains different recognition algorithms and meet the item’s needs. In order to satisfy the small size of MTD, the improved connected algorithm based on DSP was realized.By the way the CMOS image data source based on was designed. In conclusion, theimproved connected algorithm reached the requirement of real time, and get a goodrecognising result.This thesis mainly completed the pre-research of the MTD recongnise target in thespace, the thesis summed up all the work and put forward some benefit advices for thefurhter work.更多还原