欢迎访问林业科学,今天是

林业科学 ›› 2004, Vol. 40 ›› Issue (1): 154-157.doi: 10.11707/j.1001-7488.20040125

• 论文及研究报告 • 上一篇    下一篇

基于BP神经网络的树形识别系统研究

赵茂程 郑加强 林小静 封晓强   

  1. 南京林业大学,南京210037;东南大学,南京210018;
  • 收稿日期:2003-03-12 修回日期:1900-01-01 出版日期:2004-01-25 发布日期:2004-01-25

Research on Tree Crown Recognition System Based on BP Networks

Zhao Maocheng,Zheng Jiaqiang,Lin Xiaojing,Feng Xiaoqiang   

  1. Nanjing Forestry University Nanjing210037;Southeast University Nanjing210018
  • Received:2003-03-12 Revised:1900-01-01 Online:2004-01-25 Published:2004-01-25

摘要:

简要介绍了人工神经网络及分形理论的应用,提出了用分形维数特征即树木图像的灰度曲面分形维数、冠形轮廓分形维数及4个方向的有向维数和树木形状特征即树冠高度方向8等分处所对应的8个冠幅和树冠的高度作为特征来进行树形的识别,并建立了树形识别系统。该系统的输入层分别是图像的维数特征和树木形状特征值,共计15个节点,一个隐含层,输出层则是要判断的6种树形。试验表明该系统能准确地对树形进行识别。

关键词: 人工神经网络, 树形, 识别, 分形维数

Abstract:

Applications of BP network and fractional dimension were introduced. Fractional characters such as fractional dimensions of gray scale image surface and tree crown contour ,and 4 directional fractional dimensions in different directions, and shape characters of tree crown such as the height of tree crown and 8 crown breadths in 8 equal division place were employed to recognize the tree crown. And a tree crown recognition system was established in which there are 15 nodes of tree crown features and image fractional dimensions in the input layer, 6 nodes of tree crown types in the output layer and 1 node in the implicit layer. The experiment showed that the system could recognize the tree crown precisely.

Key words: ANN, Tree crown, Recognition, Fractional dimension