棉田喷药农业机器人的导航路径识别

为研究棉田农药喷洒机器人导航路径识别方法,以自然环境下采集的棉田图像为研究背景,在Lab色彩空间进行处理,把棉株从土壤背景中识别出来。通过最大方差阈值分割法将图像转化为二值图像,并经过中值滤波去除噪声。二值图像垂直方向投影做直方图,利用波谷位置确定左右垄分界线。根据左右垄棉株位置平均得到导航离散点,通过Hough变换得到导航路径,进而得到导航控制参数。利用坐标系转换关系将图像坐标系中的导航信息转换到世界坐标系,从而控制机器人行走。基于AS-R机器人对连续动态图像进行分析,该方法获得的导航参数是完全可行的。

Guidance lane detection for pesticide spraying robot in cotton fields

Natural cotton field images were analyzed in a Lab color space to study the feasibility of lane detection for agricultural robots. The cotton in the images were successfully recognized from the soil background. A maximum variance threshold and median filter were used in preprocessing to obtain binary images and remove noise respectively. Vertical histogram of the images could be divided into right and left regions, and then the guidance points were found from the average positioning information of the two regions. The best guidance lane was located using a Hough transformation which was then used to guide the machine. The parameter identification by this method is efficient and can be used to analyze large numbers of sequential images of the cotton field by an AS-R robot.