针对采摘机器人无法对重叠苹果准确分割的问题,提出了一种基于改进极限腐蚀和控制标记符分水岭分割苹果图像的方法.首先,利用果实色差分量的关系,采取R-G颜色分量作为颜色特征向量对图像进行初分割,然后采用OTSU方法分割、孔洞填充、去除小面积等方法获得完整果实二值图像,对二值图像通过改进极限腐蚀的方法来获取种子点,即局部最小区域,通过获得图像的内外标记符使用控制标记符分水岭算法,形成最终的分割图像并标记出分割线.结果表明:这种方法能够很好地改善传统方法出现错误分割的问题,找出清晰的分割线,正确分割率能够达到96.5％,较传统分水岭法和快速聚类分割算法分别提高了7.8个百分点和4.9个百分点,能够满足采摘机器人对重叠苹果图像的分割要求.%As picking robot can not segment the overlapped apples accurately, a new algorithm is proposed by im-proved limit corrosion and control segmenting apple image. R-G color component was taken as color features to segment im-age. After being segmented by OTSU method, holes filling and small area removing, the complete fruit binary image was obtained. Seed point could be obtained by improving the limit corrosion method for binary image, which was the minimum area. Segmentation image and the marked segmentation line were obtained by using the watershed algorithm. The results showed that the proposed method could solve the problem of error segmentation and find clear dividing lines. The correct segmentation rate reached 96. 5％, which improved by 7. 8％ and 4. 9％ than that of traditional watershed algorithm and fast clustering algorithm.
展开▼
