田间背景下小麦叶尖生长点提取  

Wheat tip growing points extraction in the field

作　　者：祁卫宇 王传宇[2,3] 郭新宇[2,3] 

机构地区：[1]首都师范大学,北京市100048 [2]国家农业信息化工程技术研究中心,北京市100097 [3]北京农业信息技术研究中心,北京市100097

出　　处：《中国农机化学报》2016年第8期149-154,共6页Journal of Chinese Agricultural Mechanization

基　　金：国家自然科学基金项目(31171454)

摘　　要：采用机器视觉可以实时监测作物长势,然而由于受到复杂背景和变化光照的影响,田间小麦图像叶尖生长点提取难度较大,因此本文提出一种基于深度信息进行区域生长来分割图像并提取小麦叶尖生长点的图像处理方法。首先,根据作物颜色特征去除背景,采用Canny算子检测小麦边缘,然后通过双目视觉技术,获取视差图;然后根据深度信息赋予不同的灰度值,并通过灰度阈值分割仅保留前排小麦深度图,以前排小麦深度图为种子点进行区域生长,得到前排小麦图像;最后检测小麦深度图叶尖,并将其作为初始位置,查找彩色图像前排小麦真实叶尖。结果表明该方法提取准确率为75%,能有效克服复杂背景和纹理的影响,满足应用需求,为植株生长监测提供技术支撑。Machine vision technology can be used to monitor the crop growth in real time,but it is difficult to identify the wheat leaves growing tips in field image due to the complexity of background and the variant illumination.In this paper,an image segmentation algorithm based on the depth information was presented.Firstly,image background was eliminated according to the features of crop color space distribution,and edges were detected by Canny operator.Secondly,depth information of wheat leaves was computed by binocular vision,and the wheat leaves in front row was brighter in depth map than those in back row,which could be used to carry out region growth algorithm based on edges obtained previously.At last,the wheat leaves tips in the depth map were detected,and taken as the initial position to find the true tips in the color image.The results show that the accuracy is 75%,which can effectively overcome the influence of background complexity and provide technical support for plant growth monitoring.

关 键 词：复杂纹理 深度信息 图像分割 区域生长 叶尖提取 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]