基于 Lab 空间和 K-Means 聚类的叶片分割算法研究

摘要

By classifying plant leaves has important significance in the study of plant species identification , classification in plant leaves , leaves of accurate segmentation is a necessary prerequisite to classify .This paper analyzes the contrast between the traditional threshold segmentation of the largest class clustering two variance method and K-Means segmenta-tion algorithm , to achieve segmentation leaves and RGB space conversion to Lab space , and then use two algorithms were split .The results show that the traditional threshold segmentation and K -Means clustering segmentation can not be the target image accurately segmented;in Lab space for a component of threshold segmentation can remove the shadow part , but the segmentation results for binary image;while in Lab space K-Means clustering segmentation , not only can effec-tively eliminate the shaded area in the captured image generated by the process , and after the image segmentation for col-or images , the extraction of texture and color features more convenient and improve the classification accuracy .%对植物叶片进行分类，在植物种类鉴别研究中有着重要的意义，而在植物叶片分类中，对叶片的准确分割是进行分类的必要前提。为此，对比分析了传统阈值分割中的最大类间方差法和 K－Means 聚类两种分割算法，实现对叶片的分割，并将RGB空间转换到 Lab空间，再利用两种算法分别进行分割。结果表明：传统的阈值分割和K－Means 聚类分割无法将目标图像准确地分割出来；在Lab空间对 a 分量进行阈值分割可以去除阴影部分，但是分割结果为二值图像；而在Lab空间进行 K－Means 聚类分割，不仅能够有效地消除在拍摄图像过程中产生的阴影部分，而且分割后的图像为彩色图像，对纹理和颜色特征的提取更加方便，提高了分类识别的准确率。