摘要
提出了一种基于光纤锥视觉的植物叶片检测方法,通过光纤锥与CCD传感器耦合成像获取植物叶片图像。由于叶脉复杂多变的特点,因此采用3种叶脉提取的方法,即迭代阈值法、一维信息熵法和传统的自适应阈值分割法,并用这3种方法分别对两种树叶进行叶片脉络图像提取。利用K-均值算法,针对叶片脉络图像的7个不变矩信息进行聚类,分析这3种阈值方法的优劣。实验结果表明,这3种方法都可以提取出比较理想的叶脉络和叶边缘信息,但迭代阈值法和自适应阈值法更有利于植物物种识别。
A strategy based on the optical fiber taper vision for plant leaf detection was presented in this paper , the ima-ges of plant leaves were captured by the optical fiber taper coupling with CCD .Because of the complexity of leaf vein , three techniques for extracting leaf vein were employed , they were iterative threshold , one-dimensional information entro-py detection , and traditional Otsu＇s adaptive threshold segmentation , and the three methods were applied to extract vein images of two kinds of plant leaves respectively .K-means clustering algorithm was used to classify two kinds of plant leaves based on the seven invariant moments of the vein images , to analysis the quality of the three threshold methods . The results show that all the three segmental approaches could achieve effective leaf vein and edge details , and the vein images obtained through iterative threshold and Otsu＇s adaptive threshold segmentation are more effective in plant species classification .
出处
《农机化研究》
北大核心
2013年第12期10-14,共5页
Journal of Agricultural Mechanization Research
基金
国家自然科学基金项目(51176016)
关键词
叶脉提取
光纤锥
迭代阈值
信息熵
K-均值聚类算法
vein extraction
optical fiber taper
iterative threshold
information entropy
K-means clustering algorithm
作者简介
作者简介：陈树越（1963-），男，河北定州人，教授，博士，（E—mail）Csyue2000@yahoo．cn。
