中文摘要：

植物叶片的长度是重要的叶片形态参数之一。给出了一种基于图像的长叶柄轴对称叶片中轴长度测量方法。通过扫描仪获取叶片图像;利用阈值分割和数学形态学操作提取叶片区域;利用骨架化方法提取叶片区域骨架;利用数学形态学开操作去除叶柄部分,并确定叶柄根部的骨架点作为中轴的起点;对叶面区域的骨架进行分支和交叉块的标记,并对未落在叶面区域边缘的骨架端点进行延拓;穷举式地获取所有从中轴起点到其余骨架端点的路径,并根据等面积准则和最大曲率准则确定中轴;对中轴进行折线拟合,并根据拟合折线的长度和图像的物理分辨率计算得到叶片长度。在实际叶片图像上的实验结果表明所提出的方法有效,且测量精度优于已有的、基于最小外接矩形检测的方法。

英文摘要：

Leaf length is an important morphological parameter of plant leaves. An image-based length measurement of plant leaf medial axis is proposed for axially symmetric plant leaves with elongated petioles. Leaf image is ob- tained by a scanner. A threshold followed by mathematical morphological operations gives a binary image containing a single region of the leaf. A fast skeletonization of the leaf region is performed. The petiole part is removed by a morphological opening, and the skeleton pixel at the root of the petiole region is determined as the start point of the leaf medial axis. Skeleton pixels are then grouped into clusters, between the clusters is computed. The terminal skeleton pixels and an adjacency matrix describing the connectivity that are not on the boundary of the leaf region are detected and extended, and an exhaustively enumeration process is carried out on the adjacency matrix to retrieve all the paths from the medial axis start point to the terminal skeleton pixels. Among the paths, one is determined as the medial axis via the Equal Area Criterion and the Maximum Curvature Criterion. Poly-line fitting is applied on the detected medial axis, and the length of the fitting line along with the physical resolution of the image give the result of the leaf length measurement. The experimental results on real-world plant leaf images show that the medial axes of leaves can be well detected, and the average measurement error is lower than those by the existing Minimum Enclosing Rectangle-based leaf length measurement approaches.