苗间锄草机器人信息获取方法的研究

Information acquisition method for mechanical intra-row weeding robot

该文以移栽类蔬菜用苗间锄草机器人信息获取方法为研究对象,根据锄草机器人对实时性的要求,研究了以最小耗时和最大包容准确度为目标的信息获取方法。在RGB空间内利用植物G分量值的优势去除土壤等背景干扰,根据机器人前进方向上蔬菜苗株次序排列及苗株与杂草在形态特征和分布规律上的差异,设计了图像行像素直方图的参数组合方法,确定了苗株中心沿机器人前进方向的一维坐标。试验表明,锄草机器人前进速度在0～3km/h时,苗株一维坐标检测平均误差为±5mm,算法平均耗时小于20ms。该方法能够满足系统实时检测的技术要求,为锄草机器人田间作业奠定了基础。

In order to study the information acquisition method of intra-row weeding robot for transplanted vegetables, an algorithm was developed based on fast operation speed and most accredited accuracy. Green plants could be segmented from background directly using the advantage of G component in RGB color space. As the crop seedlings grow in order and have different shape and distribution from weeds, individual crop plants could be located in the robot＇s moving direction through the histogram analysis of plant-pixels vertical to the crop rows. According to the experiment, when the robot moved with a speed from 0 to 3 km/h, the average error of the locating algorithm was ±5 mm and the average processing time was less than 20 ms. The developed method can meet the technical requirement of the weeding robot, and lay a foundation for the field work of weeding robot.