作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程

• 人工智能及识别技术 • 上一篇    下一篇

基于图的航空图像与GIS模型匹配算法

谷多玉1,申 浩1,叶曙光2,林 喆3,常红星1   

  1. (1. 中国科学院自动化研究所,北京 100190;2. 江苏金思源电气有限公司,江苏 扬州225200; 3. 北京空间机电研究所,北京 100076)
  • 收稿日期:2012-08-21 出版日期:2013-10-15 发布日期:2013-10-14
  • 作者简介:谷多玉(1984-),女,博士研究生,主研方向:视觉导航,目标检测,图像处理;申 浩,博士研究生;叶曙光、林 喆,高级工程师;常红星,研究员
  • 基金资助:
    国家自然科学基金资助项目(61005028, 61175032)

Matching Algorithm of Aerial Image and GIS Model Based on Graph

GU Duo-yu  1, SHEN Hao  1, YE Shu-guang  2, LIN Zhe   3, CHANG Hong-xing   1   

  1. (1. Institute of Automation, Chinese Academy of Sciences, Beijing 100190, China; 2. Jiangsu King Source Electric Co., Ltd., Yangzhou 225200, hina; 3. Beijing Institute of Space Mechanics & Electricity, Beijing 100076, China)
  • Received:2012-08-21 Online:2013-10-15 Published:2013-10-14

摘要: 地理信息系统(GIS)采用语义描述,图像中缺少颜色及灰度信息,只能基于其结构特征进行匹配。受目前图像处理技术限制,从图像中提取的特征信息量少、精度受限、且往往含有噪声。针对上述问题,提出一种基于图的航空图像与GIS模型匹配算法。基于UWG-SA方法分别对GIS与实时图像特征自动构建图,根据设定的相容函数,采用分级指派方法计算图的全局相容度,并给出主方向方法去除误匹配,求解变换参数。实验结果表明,该算法匹配概率为98.5%,平均匹配误差为8.54 m,平均耗时为0.075 s,可满足飞行器导航需求。

关键词: 航空图像, GIS模型, 无向加权图, 线性不变, 非精确图匹配, 分级指派, 主方向法

Abstract: Geographic Information System(GIS) is described by semantics. There is no color or gray information in GIS. The matching between aerial images and GIS has to rely on structural features. Restricted by current image processing techniques, the features extracted from images are low informative, have low precision, and usually contain noises. Aiming at these problems, a graph-based matching framework is proposed to align the GIS model and the aerial image. The Undirected Weighted Graphs(UWG) are constructed automatically using the angle between segment lines called UWG-SA for GIS and the extracted image features, respectively. A Graduated Assignment(GA) method is performed to find the global correspondences. It uses the main-direction method to eliminate the incorrect match sets and get the transformation parameters. Experimental results show that the matching probability is about 98.5%, the average matching error is about 8.54 m, and the average cost time is 0.075 s. The algorithm can basically meet the demand for aircraft navigation.

Key words: aerial image, Geographic Information System(GIS) model, undirected weighted graph, linear invariant, inexact graph matching, graduated assignment, main-direction method

中图分类号: