基于线结构光视觉的番茄重叠果实识别定位方法研究

Identification and localization of overlapping tomatoes based on linear structured light vision system

为实现番茄自动化采收,针对温室光照条件多变、背景复杂及果实簇生遮挡的特征,构建了基于线结构光视觉的番茄果实识别定位视觉系统,并提出了一种针对簇生重叠果实的区域分割方法。在(2R-G-B)色差模型下利用动态阈值分割,提取成熟果实区域;计算亮度渐变梯度向量,采用灰度重心法提取结构光条纹中心线;基于线结构光立体测量原理,在光条纹扫描视场过程中实时获取成熟果实表面三维信息,并映射形成果实表面与相机光心距离深度图像,实现簇生重叠果实区域的立体分割;基于果实表面激光条纹点空间坐标拟合果实外接球体,获取果实形态和空间坐标信息;试验结果表明,视觉系统单次识别测量耗时2.8s,果实直径测量误差小于6mm,果实与摄像机中心距测量误差小于8mm。

According to the variable illumination conditions,complex background and overlapping targets of tomato,a new method of identification and localization was proposed based on linear structured light vision system.The ripe fruit area was firstly extracted under the dynamic threshold segmentation in（2R-G-B）model.The blue line laser stripe area was then determined in B channel.The direction for image of light stripe was combined with gray barycenter method for extracting the center of laser stripe area.The ripe fruit area was then filled with gray which maped the depth information of the fruits based on the spatial localization property of the linear structured light and the difference of depth between overlapping tomatoes.The segmentation and the spatial information of the overlapping tomato were finally obtained.The field test showed that the measurement error on the fruit radius and the centre distance error between the fruit and camera were less than 6mm and 8mm respectively.