基于自适应标记分水岭算法的茶叶嫩芽图像分割方法

Image Segmentation of Tea Buds Based on Adaptive Marked Watershed Algorithm

针对茶叶嫩芽图像颜色因子差异小、难以区分的问题,提出一种基于自适应标记分水岭算法的茶叶嫩芽图像分割方法.首先对样本图片进行特征提取并分析,选定基于区域的分水岭分割算法;然后设计一种基于G通道分量的检测框,对嫩芽图片的G通道分量值进行对比计算,自适应选取最优区域中心点为起始标记点,合并梯度相似区域,以此改善传统分水岭算法在茶叶嫩芽图像上过度分割的情况;最后通过横向、纵向对比试验,验证了本文算法优于其他经典图像分割算法,更适用于茶叶嫩芽图像分割.

To address the problem of small differences in the color factor of tea bud images,which are difficult to distinguish,this paper proposes a tea bud segmentation method based on the adaptive marker watershed algorithm.Firstly,the feature of sample images is extracted and analyzed,and a region-based watershed segmentation algorithm is selected.Then,a G-component-based detection frame is designed to compare and calculate the G-component values of the bud images,adaptively select the center of the optimal region as the starting marker point and merge the gradient similar regions,thereby solving the problem of over-segmentation of the traditional watershed algorithm on tea bud images.Finally,horizontal and vertical comparison tests show that the algorithm of this paper is better than other classical image segmentation algorithms and more suitable for tea bud image segmentation.