基于最小距离求交的狭叶锦鸡儿叶片横切面宽度测量

Measurement of leaf cross-section width of Caragana stenophylla Pojark.based on minimum distance intersection

针对传统叶片横切面宽度测量效率低、劳动强度大、重现性差,已有图像处理算法与实际测量方式不匹配等问题,以狭叶锦鸡儿叶切片为研究对象,根据叶片横切面特点,提出基于最小距离求交算法实现其宽度测量。首先将采用同态滤波对叶片横切面图像进行光照补偿;接着将灰度化后图像采用最大类间方差法(Otsu)将叶片整体从背景中分离,并使用形态学处理算法,消除初分割边缘毛刺,填充存在的孔洞;然后采用Sobel算子提取边缘,使用快速行进骨架提取算法提取目标骨架;最后采用本文提出的最小距离求交法确定宽度测量点,使用欧几里得算法计算两点间图上距离,将其除以对应比例关系,得到实际叶片横切面宽度。实验结果表明:对供试叶片宽度测量均值约为646.31μm,均方根误差RMSE约为3.065,相较ToupTek Toupview软件交互测量,相对误差约为0.42%,满足实际测量需求。

In view of the problems of low measurement efficiency,high labor intensity,and poor reproducibility of traditional blade cross-sectional widths,existing image processing algorithms and actual measurement methods do not match.In this paper,Caragana Stenophylla Pojark.leaf slices are used as research objects.According to the characteristics of blade cross-section,a crossover algorithm based on minimum distance is proposed to achieve its width measurement.Firstly,a dynamic Butterworth filter is used to homomorphically filter the cross-sectional image of the blade to eliminate uneven illumination of the microscopic image;then the whole blade is separated from the background by using the Maximum Between-Class Variance method(Otsu)in the grayscale image,use the morphological opening operation to eliminate edge burr and small connection,and use the morphological reconstruction algorithm to fill the segmented holes;then use the Sobel operator to extract the edge,use the fast-moving skeleton extraction algorithm to extract the target skeleton.Finally,use the minimum distance intersection method proposed in this paper to determine the width measurement point,and the Euclidean algorithm is used to calculate the distance between the two points on the map,which is divided by the corresponding proportional relationship to obtain the actual blade cross-sectional width.The experimental results show that the average measurement of the blade width of the test is about 646.31μm,and the RMSE is about 3.065.Compared with the interactive measurement of ToupTek Toupview software,the relative error is about 0.42%,which meets the actual measurement need.