基于机器视觉的砧木定位识别方法

摘要

Machine vision and image processing technology was used in order to determine accurate connection point of a parental stock on a rootstock and recognize cotyledon angle. Morphological principle was applied to process the image and obtain the feature of rootstocks. Eroding process was used to remove the euphylla and get the angle between the horizontal and cotyledons, thus obtain the adjusting angle for connection. Results showed that the maximum absolute error of the connection point was 2. 09 mm, while the mimimum was 0. 03 mm. The accuracy of position recognition of cotyledons and euphylla was 97. 9%.%为了实现蔬菜嫁接过程中砧木抓取点和子叶方向的精确定位,对机器视觉采集的砧木苗图像进行处理.对砧木侧视图进行区域分割判断子叶与茎的连接处,作为抓取点；对砧木俯视图利用形态学处理分离子叶和真叶,根据区域长短轴差值识别真叶并予以去除,计算子叶与水平线的夹角,得出旋转角度.结果显示,基于机器视觉方法的砧木抓取点的判断,绝对误差最大为2.09 mm,最小为0.03 mm;利用机器视觉判断砧木旋转角度,真叶和子叶区分的成功率达97.9％.