作　　者： ; ; ; ;

机构地区： 中国农业大学工学院

出　　处： 《中国农业大学学报》 2013年第1期165-171,共7页

摘　　要： 为精确识别和定位玉米行间杂草,满足基于机器视觉的变量施药系统喷施要求,提出了一种基于行宽的多行玉米行间杂草识别算法。该算法以垂直拍摄的3叶期3行玉米田间图像为研究对象,利用YIQ颜色空间中的Q分量灰度化田间彩色图像,以降低自然光源对图像的影响;通过建立实际田间玉米行宽与图像玉米行宽的映射关系,将3叶期玉米行的宽度映射到对应图像中,并确定基于识别率和运算速度的覆盖范围;以具有一定宽度的玉米行作为识别基准,减小未连通叶片区域的误识别率,提高对杂草识别的精度。从识别精度和速度2方面与基于作物行中心线识别算法进行了对比。研究结果表明,对于3叶期3行玉米田间图像,杂草正确识别率可达89.2%,速度为197ms。本算法有效地提高了行间杂草识别的精度和速度,能够初步满足基于机器视觉的变量施药系统对大田玉米多行喷施的工作要求。 in order to meet the requirement of variable rate pesticide spray based on machine vision by detecting the weeds among the corn rows, this paper presents an identification algorithm of weeds among multi-row corn based on rows＇ width. After taking the image of 3-leaf stage corn vertically down to the crops over three corn rows, Q color component was extracted through Otsu method to reduce the impact of natural light. The span of 3-leaf stage corn rows in the image was determined by establishing the mapping of the graphical width with the corresponding actual corn row width. The detection of the weeds depending on the corn row mapping could reduce the false identification rate of the patch unconnected with the row and improve the recognition accuracy of the weeds. The weeds identification algorithm among multi-row corn was compared with crop centerline detecting algorithm in this paper based on two aspects of identification accuracy and speed. The results show that the weed identification rate and speed can amount to 89.2% and 197ms, respectively. This means the algorithm can fairly match the variable rate pesticide spraying based on machine vision system.

关 键 词： 玉米 杂草识别 行宽 颜色空间 变换

领　　域： [自动化与计算机技术] [自动化与计算机技术] [农业科学]