基于最小相切圆和形态学相结合的作物行检测算法

机器视觉的农业导航路径规划是精准施药的关键,而作物行提取是其准确识别作物行路径的基础。为此,以玉米为研究对象,提出了一种基于最小相切圆原理和形态学相结合的作物行检测算法。首先在室外田间环境下采集生长早中期的玉米作物行图像,选择作物行比较规整的图像进行处理;其次,利用改进的超绿灰度化(1.8R-G-B)算法对玉米作物行图像进行灰度化处理,大大减少了噪声的干扰,通过中值滤波基本消除了噪声;然后,运用Otsu阈值算法获取了玉米作物行的二值图像。由于作物行呈线型,在此基础上,采用5×1像素的线型结构元素和3×3像素的方形结构元素两者相结合的方法对二值图像进行腐蚀、膨胀运算,并采用提出的最小相切圆与形态学结合的方法提取中央玉米作物行的骨架并进行中央作物行直线的拟合。实验表明:该算法能提供准确的位置信息,且对作物行边缘噪声具有较强的抗干扰能力,对进一步研究精准施药提供了参考依据。

A Crop Row Detetcion Algorithm Based on Minimum Tangent Circle and Morpholo gy

Path planning in agricultural machine vision navigation is the key to accurate spraying, but the crop rows de-tection is the basis for accurate identification of crop rows path.Taking corn as the research object,the paper propose a crop row detection algorithm based on minimum tangent circle and morphology.First,collected corn crop image growing early and middle time in the outdoor environment, choose the crop rows that more structured to process, and secondly,u-sing the improved 1.8R-G-B algorithm to realize gray transformation,greatly reducing the noise and by median filtering eliminate the noise virtually.And then use the Otsu threshold algorithm to obtain a binary image of the corn crop rows. Because crop rows is linear,on this basis, using the method that combine linear structure element 5 ×1 pixel with square structure that 3 ×3 pixel to conduct erosion and dilation.Finally,we use this method based on minimum tangent circle and morphology to detect central line skeleton and fitting the line of the crop.Experiments show that the algorithm can provide accurate location information, and to the noise on the edge of the crop rows has a strong anti-jamming capa-bility, provides reference for further research in precision pesticide.