摘要
利用计算机视觉技术将杂草从农作物和土壤中区别开来已成为精细农业领域研究的热点问题。提出了一种颜色和形态特征相结合的杂草实时识别方法。在YcbCr颜色模型中,以色差Cr为特征量、以最大类间方差作为GA的适应度函数对Cr进行自适应阈值分割将植物与背景分离;利用植物的形态特征,结合形态学腐蚀、膨胀方法及差影法将农作物和杂草分离。多幅杂草图像研究结果表明:该算法杂草正确识别率大于83.1%,处理一幅640像素×480像素的图像平均只需38ms,识别速度满足25帧/秒的实时性要求。
The technology of weed recognition based on machine vision becomes the research focus of precision agriculture. In order to realize the variable sparing technology, it is required to identify weed firstly. This paper presents a weed identification method that combines color and morphological features. Color feature is utilized to distinguish plants and background： using a method that takes YCbCr as color-space and Cr as characteristic variant and maximum variance as criterion. Morphological feature is utilized to distinguish crops and weed： using a method that combines morphological filter and differential shadow algorithm. Experiments on a serial of weed images are conducted. The experimental results show that the correct identification ratio exceeds 83.1%, the processing speed of a 640 pixels × 480pixels image is about 38ms and the identification soeed exceeds 25fps.
出处
《光电工程》
EI
CAS
CSCD
北大核心
2006年第7期96-100,共5页
Opto-Electronic Engineering
关键词
杂草识别
机器视觉
目标识别
形态学
Weed recognition
Machine vision
Target recognition
Morphological
