ISSN 1000-1239 CN 11-1777/TP

计算机研究与发展 ›› 2015, Vol. 52 ›› Issue (6): 1409-1423.doi: 10.7544/issn1000-1239.2015.20131422

• 图形图像 • 上一篇    下一篇



  1. (东北师范大学计算机科学与信息技术学院 长春 130117) (智能信息处理吉林省高校重点实验室(东北师范大学) 长春 130117) (
  • 出版日期: 2015-06-01
  • 基金资助: 

A Terrain Skeleton Feature Extraction Method Based on Morphological Encoding

Zhang Huijie, Liu Yaxin, Ma Zhiqiang, He Xinting, Bao Ning   

  1. (School of Computer Science and Information Technology, Northeast Normal University, Changchun 130117) (Key Laboratory of Intelligent Information Processing of Jilin Universities (Northeast Normal University), Changchun 130117)
  • Online: 2015-06-01

摘要: 鉴于剖面识别骨架特征提取方法(profile recognition and polygon breaking algorithm, PPA)准确性差、不能提取特殊地形等问题,提出了一种结合形态学的剖面识别骨架特征提取方法(profile recognition and polygon breaking algorithm based on morphology erosion, MEPPA).通过剖面识别提取原始的骨架特征候选点,根据方向系数连接成多边形条带,在此基础上提出了生成标量特征域的填充算法;引入形态学区域细化思想,提出了形态编码和骨架特征形态简化算法,将特征域简化为骨架特征线;为了满足各领域对矢量骨架特征的需求,提出了标量特征线复原、检测与优化剔除等策略,准确地复原了矢量骨架特征模型;提出了保留外分支和环路特征的解决方案,解决了传统骨架特征提取方法不能保留较长的主干线以及不能提取环路地形特征的问题.在真实数据上进行了实验研究,结果表明提取的骨架特征形态的整体效果优于传统方法.

关键词: 特征候选点, 标量特征域, 形态学, 剪枝, 目标点识别

Abstract: Since the current profile recognition methods are not able to extract the precise skeleton and the special terrain feature, a new profile recognition method combined with morphology is proposed to solve these problems. In this method, the candidate points are extracted by the profile recognition, and then are connected into polygon stripe according to direction coefficients. Furthermore, the fulfill algorithms building the scale feature areas are put forward based on the polygon stripe by taking advantage of morphological strategy. In addition, multiple morphological codes are proposed to simplify the scale feature areas according to the concept of morphological erosion algorithm, and to obtain the scale feature lines. In order to satisfy the requirements about vector skeleton features in the various fields, the algorithms including restoration, detection and optimization are proposed to implement the transformation from the scale model to the vector model. Finally, a series of strategies about keeping the out-branches and recognizing the ring process are presented in this paper, which solve the problems about missing the long trunk lines and ring features in the result feature lines. These methods have been tested on the benchmark data and the real elevation data. As a result, the skeleton feature lines produced by our method outweigh the traditional method as a whole.

Key words: feature candidate point, scale feature area, morphology, pruning, target point recognition