实时识别行间杂草的机器视觉系统

在实验室环境条件下,开发和测试了识别行间杂草的机器视觉系统。硬件系统主要由速度可控的土壤箱设备、三台实时采集图像的摄像机和计算机组成;软件系统根据植物和背景的颜色特征二值化图像,再根据田间作物的位置特征识别作物和行间杂草。实验表明,采集并处理一幅大小为710×512像素的彩色图像的平均时间为426 ms,系统的正确识别率达到了86%。

Machine vision system used for real-time detection inter-row weeds

A machine vision system to detect inter-row weeds was developed and tested in the lab. The hardware system was mainly made up of a soil-bin device with controllable velocity carriage, 3 CCD cameras used to capture the images and a PC. The software system was developed to transform color images to binary images by the color feature of plants and background, and to distinguish crops and inter-row weeds by the location feature of crop within the field. It indicated that the mean of executed time of capturing and processing a color image (710×512 pixels) was 426 ms, and the correct classification rate of the system was 86%.