成熟期棉田棉花行精准识别算法研究

Research on Accurate Recognition Algorithm of Cotton Row in Cotton Field in Mature

为满足采棉机器人成熟期棉花的收获要求,提出一种成熟期棉花行中心线快速准确的检测方法。在自然环境下相机采集成熟期棉花图像,利用灰度变换、高斯滤波等方法对图像进行预处理,其次通过标记分水岭分割算法对图像进行分割,滤除图片中杂点,将棉花信息从背景信息中提取出来。最后通过像素坐标分割法检测棉花特征点并采用最小二乘法拟合棉花行中心线。实验结果表明,像素坐标分割算法在不同光照情况下、棉花种植稀疏与稠密情况下都能够准确的拟合出棉花行走向的中心线,速度快,准确率高,可以满足采棉机器人实时作业要求。

In order to meet the cotton harvesting requirements of the cotton picking robot in the mature period,a rapid and accurate detection method for the center line of the mature cotton row is proposed.Firstly,gray transformation and Gaussian filtering are used to preprocess the cotton image collected by the camera in the natural environment.Secondly,the marked watershed segmentation algorithm is used to segment the image,filter out the clutter in the image,and extract the cotton information from the background information.Finally,cotton feature points are detected by pixel coordinate segmentation method and cotton row centerline is fitted by least square method.The experimental results show that the pixel coordinate segmentation algorithm can accurately fit the center line of the cotton walking direction under different lighting conditions and under the conditions of sparse and dense cotton planting.The speed and accuracy are high,which can meet the real-time operation requirements of cotton picking robots.