基于机器视觉的自然环境下作物行识别与导航线提取

Crop Recognition and Navigation Line Detection in Natural Environment Based on Machine Vision

针对基于机器视觉的农业导航机器人在图像处理时易受光照变化影响和常规导航线检测算法实时性、稳健性不高等问题,提出了YCrCg颜色模型,选择该颜色模型中与光照无关的Cg分量进行后续图像处理,采用基于二维直方图的模糊C均值聚类法(FCM)进行图像分割,并根据图像中作物行的特点,提出了基于直线扫描的作物行直线检测算法。该算法将图像底边和顶边像素点作为直线的两个端点,通过移动上下端点位置产生不同斜率直线,选择包含目标点最多的直线作为作物行中心线。实验表明,不同光照下基于YCrCg颜色模型的图像分割可以有效地识别出作物行,处理一幅640pixel×480pixel图片耗时约为16.5ms,直线扫描算法能快速准确的检测出导航线,与最小二乘法、Hough变换等算法相比具有速度快、抗干扰性强等优点。

In order to solve the problems of serious illumination interference for image processing and the poor robustness of conventional navigation line detection algorithms in agricultural navigation robot based on machine vision, the methods of crop recognition and navigation line extraction in natural environment are studied. Cg component of YCrCg color model is selected for subsequent image processing to reduce the adverse effects of light change on image segmentation and navigation line extraction. The fuzzy C-means clustering method （FCM） based on two-dimensional histogram is used for Cg component segmentation, so as to identify the gree^crop. According to the characteristics of crop rows in mage, a method of crop line detection based on linear scanning is designed. Pixel on image bottom and top edge are selected as two endpoints of a straight line, by moving the endpoints location result in different slope lines, the line containing the most target points is chosen as the crop centerline, and then obtain the navigation line. The experimental results show that image segmentation based on YCgCr color model can effectively identify the crops under different illumination conditions. Further more, the time consumption for single image of 640 pixel× 480 pixel is about 16.5 ms. The linear scanning algorithm can quickly and accurately find the navigation line. Compared with Hough transform and least square algorithm, the designed algorithm has the advantages of high speed and good robustness.