基于分水岭算法的柑橘识别技术研究

Citrus recognition technology based on watershed algorithm

针对自然环境下柑橘粘连重叠、难以识别的问题,提出一种基于分水岭算法的柑橘快速识别方法。将采集的RGB图像转换至HSV空间,分别对H、S通道图像进行二次迭代阈值分割将柑橘目标从枝叶背景中提取后,应用分水岭分割算法对粘连重叠的柑橘进行轮廓分离,最后使用最小外接圆拟合柑橘轮廓,提取圆心坐标和外接圆半径,实现对柑橘的识别。

This paper proposes a rapid identification method of citrus based on a watershed algorithm,which solves the problem of overlapping and difficult identification of citrus in the natural environment.First,RGB images were converted to HSV space.Second,images in H and S channels were segmented with the quadratic iteration threshold segmentation method to extract citrus targets from the foliage background.Then watershed segmentation algorithm was used to separate the citrus contour.Finally,the minimal circumscribed circle was used to fit the citrus contour and the center coordinates and radius of the circle were extracted,then recognition of citrus was realized.