模糊超像素分割算法的无人机影像烟株精细提取

Tobacco fine extraction from UAV image based on fuzzy-superpixel segmentation algorithm

实现烟草单株自动化提取对加快烟草农业信息化有着重要意义,但目前烟株精细的提取还存在较大困难。因此,提出了一种基于模糊超像素分割(fuzzy-superpixels,FS)算法的无人机烟株提取方法。首先通过绿地提取方法得到无人机影像中的植被覆盖区域;然后利用FS算法对影像进行超像素分割,并统计超像素的均值、亮度、形状指数、长宽比、自定义植被指数等特征;最后通过计算超像素的特征阈值,对烟株数目进行精细提取和统计。选取3景无人机影像作为实验数据,实验结果表明,该方法提取烟株的总体精度分别为84.28%,89.05%和82.97%。该方法可实现对烟株的自动化提取,为后期计算烟草产量提供有效参考。

The successful extraction of tobacco single plant automation is of great significance to the realization of tobacco agricultural information,but there are still great difficulties in tobacco fine extraction.Therefore,a tobacco extraction method based on Fuzzy superpixels(FS)algorithm is proposed.Firstly,vegetation coverage area in UAV image is obtained by green space extraction method;secondly,super-pixel segmentation of image is carried out by using FS algorithm,and the mean value,brightness,shape index,aspect ratio,custom vegetation index and other features of super-pixel are counted;finally,the number of tobacco plants is extracted and counted by calculating the feature threshold of super-pixel.Three UAV images were selected as the experimental data.The experimental results show that the overall accuracy of this method is 84.28%,89.05%and 82.97%respectively.This method can be used for automatic extraction of tobacco plant and can provide effective reference for later calculation of tobacco yield.