基于视觉伺服的温室番茄植株主茎跟踪与测量方法  

Tracking and Measuring Method of Tomato Main-stem Based on Visual Servo

作　　者：冯青春[1,2] 王秀[2,3] 刘继展 成伟[2,3] 陈建 FENG Qingchun;WANG Xiu;LIU Jizhan;CHENG Wei;CHEN Jian(College of Engineering,China Agricultural University,Beijing 100083,China;Beijing Research Center of Intelligent Equipment for Agriculture,Beijing 100097,China;Beijing Key Laboratory of Intelligent Equipment Technology for Agriculture,Beijing 100097,China;School of Agricultural Equipment Engineering,Jiangsu University,Zhenjiang 212013,China)

机构地区：[1]中国农业大学工学院,北京100083 [2]北京农业智能装备技术研究中心,北京100097 [3]农业智能装备技术北京市重点实验室,北京100097 [4]江苏大学农业装备工程学院,镇江212013

出　　处：《农业机械学报》2020年第11期221-228,共8页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金项目(61703048);北京市农林科学院青年科研基金项目(QNJJ201722);江苏大学农业装备学部项目(4111680002)。

摘　　要：根据温室番茄智能管理作业视觉信息获取需求,研究了番茄植株主茎动态跟踪与立体测量方法,以提高对叶、果和花等目标的搜索效率。结合工厂化番茄种植特征,采用二自由度双目云台摄像机采集植株主茎图像;在对摄像机与旋转云台之间坐标关系进行标定的基础上,提出针对番茄植株主茎图像跟踪采集的云台伺服控制方法,对作业区域内植株进行自下而上多视角图像动态采集;对相邻视场主茎重叠区域的图像匹配方法进行研究,实现了植株离散图像的拼接和形态恢复;基于主茎跟踪参考点的空间坐标信息,研究了作业区域主茎长度、高度和生长倾角等立体形态参数的测量方法;最后,通过现场试验对主茎拼接与测量方法进行验证。结果表明,在距地面高度600~1500 mm作业区域内,视觉系统跟踪采集的主茎3个区域图像的平均拼接偏差为3.77°;以人工测量结果为对照,采用视觉系统测量主茎长度、高度和生长倾角的决定系数分别为0.9933、0.8426、0.9793,平均测量偏差分别为46.20 mm、18.60 mm和4.33°。本研究可为番茄智能化整枝、采摘和授粉等作业视觉信息获取提供技术支撑。Aiming at target's visual information acquisition for robotic management in tomato greenhouse,the method of tracking and measuring the plant's main-stem was researched,which was supposed to improve the detection efficiency on the object,such as fruit,leaf and flower.According to the tomato's factory-planted condition in greenhouse,a binocular vision system with two-freedom pan-tilt mechanism was adopted to capture plant's image and the calibration on the relationship between camera and pan-tilt coordinates data was introduced.The pan-tilt's servo control method for tracking the plant's main-stem was proposed,so that the camera could automatically capture the plants bottom-up images.The image matching method for splicing the discrete main-stem in the adjacent view-fields was researched,so as to recover the plant's image morphology.Based on the 3D coordinate data of a series of tracking reference points,the main-stem's length,vertical height,and inclination angle could be estimated.Finally,the method for tracking and measuring tomato's main-stem was tested in greenhouse.As the result showed,in the working area from 600 mm to 1500 mm high from the ground,the vision system could capture three images of various view-fields,and the main-stem's average splicing deviation was 3.77°.Compared with manual measurement results,the results of the automatic measurement method on main-stem's length,vertical height and inclination angle,respectively had the determination coefficients of 0.9933,0.8426 and 0.9793,and the average deviations of 46.20 mm,18.60 mm and 4.33°,respectively.In view of the performance on tracking and measuring target,the method was expected to be a support for researching on tomato's automatic pruning,harvesting and pollinating.

关 键 词：番茄植株 主茎形态 农业机器人 视觉伺服 立体测量 

分 类 号：S641.2[农业科学—蔬菜学] TP242.6[农业科学—园艺学]