基于三维点云的蔬菜大棚杂草识别方法

为实现蔬菜大棚内自动化除草的目的,针对其中的杂草识别环节,提出了一种基于三维点云的新型蔬菜大棚杂草识别方法。采用RGB-D相机获取青菜田、生菜田的三维点云图像,采用超绿色算法去除其中的土壤等背景,采用体素滤波法在保留点云图像形状特征的同时降低点云数量,然后采用欧式聚类法分割出单株青菜和单棵杂草的点云簇,分别计算得到每个点云簇的最高点的Z坐标值,最后结合深度信息Z坐标值实现蔬菜大棚杂草识别。结果表明:这种基于三维点云的杂草识别方法能够有效的识别出杂草,识别率为86.48%。该方法能够对蔬菜大棚中的杂草进行准确识别,为蔬菜大棚自动化除草提供有效的解决方案。

A Weed Identification Method in Vegetable Greenhouses Based on Three-dimensional Point Cloud

In order to realize the purpose of automatic weeding in vegetable greenhouses,a new method for weed identification in vegetable greenhouses based on three-dimensional point cloud is proposed for the weed identification.This method used the RGB-D camera to obtain three-dimensional point cloud images of green vegetable fields and lettuce fields.We used the improved super green algorithm to remove backgrounds such as soilinit,voxel filtering method was used to reduce the number of point clouds while retaining the shape characteristics of the point cloud image,then we used the European clustering method to segment the point cloud clusters of a single vegetable and a single weed in the pre-processed point cloud image,and the Z coordinate values of their highest points were calculated.Finally,the Z coordinate value of depth information was combined to realize the vegetable greenhouse weed identify.The results showed that the weed recognition method based on three-dimensional point cloud could effectively identify weed,and the recognition rate was 86.48%.The method could accurately identify weed in the vegetable greenhouse,and provide an effective solution for the automatic weeding of the vegetable greenhouse.