评价通过机器视觉的方法进行穗行数自动检测的可行性,提出一种基于边界标记和离散曲率的边缘算法确定行数,建立了检测模型。在扫描78幅不同玉米品种的果穗横断面图像的基础上进行检验。结果表明:试样玉米的穗行数为12～18行,检测的绝对误差为0.103行,相对误差率为0.66%,零误差率达到90%以上。实验证明检测算法真实可靠,机器视觉具有成本和速度上的优势,能够用于玉米穗行数的自动检测。 To evaluate the feasibility of automatic detection of ear number by machine vision method, an edge algorithm based on boundary markers and discrete curvature is proposed to determine the number of rows and a detection model is established. The cross-sectional images of 78 ears of different maize varieties were scanned. The results showed that the number of ears per sample was 12-18 rows, the absolute error of detection was 0.103, the relative error rate was 0.66%, and the zero error rate was over 90%. Experiments show that the detection algorithm is true and reliable, machine vision has the advantages of cost and speed, and can be used for the automatic detection of the ear number of maize.