基于区域生长的采摘机器人视觉识别方法

发布时间：2018-05-21 10:21

  本文选题：图像分割 + 区域生长　； 参考：《农机化研究》2017年03期

【摘要】：提出了一套基于茄子图像的空间位置信息和颜色因子相融合的区域生长分割算法。为保证茄子图像分割最佳的颜色空间和颜色因子,提取了50幅不同光照条件下的茄子图像的RGB颜色空间分量灰度图和直方图,比较了茄子果实、叶子、茎秆和空隙等的颜色特征,得出了G-B颜色因子对于茄子果实分割最为有利的结论。按照灰度级相同和空间8邻域连通的原则确定种子区域,进而通过扫描整幅图像进行初始分割。融合G-B颜色因子和空间信息对初始区域进行合并,直到分割形成的区域类间距离最大时停止生长。通过顶点链码与离散格林技术提取出果实的最小外接矩形,求解果实的生长位姿,试验表明:其分割效率均大于93%,平均用时为0.32 s,能够满足果蔬采摘机器人对视觉系统的要求。
[Abstract]:A regional growth segmentation algorithm based on spatial location information and color factor fusion based on Eggplant image was proposed. In order to ensure the best color space and color factor of eggplant image segmentation, the gray map and histogram of RGB color space component of eggplant images under 50 different illumination conditions were extracted, and the fruit, leaf and stem of eggplant were compared. The color characteristics of the stalk and the gap, and the conclusion that the G-B color factor is most favorable to the eggplant fruit segmentation is drawn. The seed region is determined according to the principle of the same gray level and the space 8 neighborhood connectivity, and then the initial segmentation is carried out by scanning the whole image. The initial region is merged with the G-B color factor and the spatial information, until the segmentation is divided. The minimal outer rectangle of fruit was extracted by vertex chain code and discrete Green technique, and the growth position of fruit was solved by the vertex chain code and discrete Green technique. The experimental results showed that the segmentation efficiency was more than 93% and the average time was 0.32 s, which could meet the requirements of the visual system of fruit and vegetable harvester.
【作者单位】： 山东科技大学机械电子工程学院;潍坊学院机电与车辆工程学院;
【基金】：国家自然科学基金项目(51505337) 山东省自然科学基金项目(ZR2014EEP013)
【分类号】：S225;TP391.41
，

本文编号：1918812