摘要: |
【目的】提出一种基于面向对象图像特征提取的植物叶面积测量方法,为快速、高精度地测量野外采集植物的叶片面积提供支持。【方法】以扫描图像为基础,借鉴遥感影像的面向对象图像特征提取的思想,获得扫描对象的矢量轮廓,以此计算其面积,并采用AutoCAD绘制的7种多边形进行重复试验,以验证该方法的精确性;然后进一步对青蒿(Artemisia carvifolia)、臭蒿(Artemisia hedinii)、苜蓿(Medicago sativa)3种植物叶片进行重复试验,并与矢量化方法、监督分类方法进行对比,分析该方法在实际叶片测量中的稳定性和计算效率。【结果】利用基于面向对象图像特征提取的植物叶面积测量方法,在进行标准几何图形的面积测量时,该方法的相对误差皆小于1.86%;与矢量化方法、监督分类方法相比,该方法在测量真实植物叶片面积时具有更高的稳定性,而且耗时都小于20 s,用时最短;该方法采用IDL模块设计,可实现叶片面积的自动批量处理。【结论】基于面向对象特征提取的植物叶片面积测量方法,叶片面积高精度及批量自动化测量的一种新途径。 |
关键词: 叶面积测量 面向对象 特征提取 |
DOI: |
分类号: |
基金项目:国家自然科学基金项目(41271297) |
|
Measuring plant leaf area based on object-oriented feature extraction |
LIU Shicheng,WEN Zhongming,QI Dehui,et al
|
Abstract: |
【Objective】An object-oriented feature extraction method was proposed to quickly and accurately measure area of plant leaf sampled from field.【Method】The scanned images of objects were collected and vector edge of objects was extracted by the object-oriented feature extraction method used in remote sensing to calculate the area.The accuracy was tested with repeated trials using seven standard geometrical objects in AutoCAD.Additionally,the areas of real leaf samples of Artemisia carvifolia,Artemisia hedinii,and Medicago sativa were measured to analyzed the stability and efficiency compared to unsupervised classification and vectorization methods.【Result】The object-oriented feature extraction based method had the relative error of the less 1.86% in measuring standard geometrical objects.Compared with unsupervised classification and vectorization methods,the established method was more stable and used less time (<20 s) in measuring real plant leaves.This method can achieve batch processing of leaves with IDL design.【Conclusion】The object-oriented feature extraction based method provides a new and accurate approach for indoor batch processing of plant leaf area. |
Key words: measurement of leaf area object-oriented feature extraction |