Abstract: 【Objective】 In order to solve the problem of rapid identification of cotton leaf diseases under complex background in cotton field, it is very important to extract cotton plants and segment cotton leaves. Therefore, this paper proposes a fast method of extracting cotton plants and segmenting cotton leaves. 【Method】Firstly, the cotton plant was separated from soil by RGB color feature of cotton plant, and then the stem of cotton plant was removed by combining morphological processing technology and color segmentation method. Then, the cotton leaf image was segmented and extracted by breadth search segmentation algorithm, watershed segmentation algorithm and contour search segmentation algorithm, respectively. 【Result】【Conclusion】The results showed that the watershed segmentation algorithm based on distance transformation had the problem of over-segmentation. The combination of breadth-first search and edge detection had a significant effect on image segmentation with clear blade structure, but compared with contour search algorithm, the latter had a wider applicability, clear contour hierarchy structure and the best segmentation effect.

Key words: image segmentation; breadth search; watershed algorithm; edge detection; contour extraction

摘要： 【目的】解决棉田复杂背景下棉花叶部病害快速识别问题,为提出一种快速提取棉花植株和分割棉花叶片的方法。【方法】通过棉花植株的RGB颜色特征将植株与土壤进行分离,结合形态学处理技术和彩色分割方法,将植株的茎秆去除,保留棉花叶片部分;分别使用广度搜索分割算法、分水岭分割算法和轮廓搜索分割算法,对棉花叶片图像进行分割提取。【结果】基于广度搜索的分割算法将叶片轮廓搜索出来与原图融合分离出叶片,该算法对于图像结构简单的情况分割效果较好,通过对应的drawContours函数将每个轮廓画出,再与原图定位,将叶片的完整信息也轮廓结合,实现叶片的分离。【结论】基于距离变换的分水岭分割算法存在过分割问题,基于广度搜索分割算法与边缘检测结合对于叶片结构清楚图像分割效果显著,与轮廓搜索算法相比,后者的适用性更广,提取的轮廓层次结构也清楚,分割效果最佳。

关键词: 图像分割, 广度搜索, 分水岭算法, 边缘检测, 轮廓提取