针叶苗木计算机视觉特征提取方法

为解决基于计算机视觉技术的针叶树种苗木自动分级系统的关键技术问题，提出了一种应用精确定位的苗木图像特征标志点提取苗木计算机视觉特征的新方法。该方法通过分析苗木的形态特征。利用苗木图像灰度矩阵和背景分离的灰度矩阵分别得到了５个特征标志点和苗木轴向边缘标志点矩阵。并在标志点的基础上，应用相应的计算方法实现了苗高、地径、根长、冠投影面积、根投影面积等１２个苗木视觉特征量的提取，为苗木自动分类器提供了有效的特征输入。实验结果表明，本方法具有计算速度快、易于实现、测量准确等优点。

Study on Extraction of Computer Vision Feature of Conifer Seedling

In this paper , in order to solve the key technological problem in conifer seedling classification system based on computer vision, a new method to extract features of conifer seedling by accurately locating feature mark points of seedling image is discussed. With carefully analyzing the morphological feature of seedling, the authors get five feature mark points and the feature point of seedling axial direction edges from two gray level matrixes. One matrix is formed by gray values of seedling image and the other is formed by gray values of the same image which is removed the background. Based on these feature mark points, twelve vision features of seedling. such as shoot height, root collar diameter, root length, shoot projected area, root projected area and so on, are extracted through corresponding calculation methods. Consequently effective feature inputs are supplied to the auto-classifier of conifer seedling. The experimental results show that this method calculates rapidly, measures accurately and is prone to be carried out.