一种新颖的分割算法在果业机器人视觉中的应用(英文)

Application of a Novel Segmentation Algorithm to Robot Vision of Fruit Industry

传统流域变换主要依赖图像梯度,该方法并不完善。在噪声图像中,边缘提取也不能够达 到很好的效果。提出的多尺度流域变换算法主要利用一组结构元素按照一定顺序,进行腐蚀膨胀的迭代来分割灰度图像。在多尺度测地重建滤波下,不同尺度的梯度流域分割线存在严格的因果关系。实验结果证实多尺度流域分割的性能远比传统方法优越。

The traditional watershed transform mainly depends on image gradient but this method is imperfect. The edge extraction from noise image cannot obtain a very good effect. The gray image segmentation is mainly carried out by using a group of structural elements to iterate corrosion expansion according to a given sequence in the proposed multi-scale watershed transform algorithm. The strict causality exists in the different scales of gradient watershed segmentation under the multi-scale geodesy reconstructing filtering. The experimental results demonstrate that the performances of multi-scale watershed segmentation are far superior to that of the traditional methods.