基于图像几何特征的行星表面参考区域提取与匹配方法  

Autonomous Referenced Area Extraction and Matching on Planet Surface Based on Geometrical Characteristics of Image

作　　者：李建更[1] 贾真[1] 阮晓钢[1] 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100124

出　　处：《北京工业大学学报》2014年第7期1066-1072,共7页Journal of Beijing University of Technology

基　　金：国家"973"计划资助项目(2012CB720000)

摘　　要：针对探测器在行星表面精确着陆的问题,提出了一种通过提取参考区域光照阴影及其轮廓几何特征进行自主识别、匹配的方法.首先,对图像预处理,包括基于区域斜分的二维最大熵双阈值分割、提取连通区域等操作;然后,提取特征区域阴影的轮廓,针对区域阴影及其轮廓信息进而提取其线状及面状几何特征;最后,利用基于加权欧氏距离的聚类方法进行模板匹配.该方法克服了在以往研究中选取参考区域单一化的缺陷,通过对取样图像大量反复的实验,匹配的正确率在93%以上,从而验证了该方法的可行性和鲁棒性.According to the precise landing of the detector on planet, an algorithm for autonomous extraction and matching of referenced area based on the geometrical characteristics of image was proposed. Firstly, the sample image was preprocessed, including threshold segmentation and connected area extraction. Then, the outline of referenced area＇s shadow and its linear and planar geometrical characteristics were extracted. Finally, template matching was tested based on Euclid distance with weights. The proposed algorithm could overcome the defect that the selected referenced area was simple. Experiments demonstrate that the accurate rate of this algorithm is higher than 93% , which proves that the proposed method is feasible and robust.

关 键 词：行星软着陆 自主光学导航 双阈值分割 几何特征 加权欧氏距离 模板匹配 

分 类 号：V448.233[航空宇航科学与技术—飞行器设计]