基于Matlab7.0的大豆苗期土壤背景与杂草分割算法

Algorithmic Division at Soybean Seedling Stage of the Soil Background from Weeds Based on Matlab7.0

解决大豆苗期图像中的土壤背景分割是大豆田间杂草识别的前提和基础。为了解决大豆苗期田间杂草识别中受光照影响及环境适应性差等问题,通过对400幅不同环境下苗期大豆图像的植被颜色和背景颜色分量的统计分析,得出用3基色红(R)绿(G)蓝(B)合成的同一像素点,绿色植被图像的颜色分量G值都大于R和B值,而背景则恰恰相反。研究表明,采用G-R和G-B双阈值颜色特征分割进行土壤背景分离取得了很好的效果,此方法较2G-R-B颜色特征分割法对绿色植物和土壤背景的分离更为有效,可广泛应用于处于各种农作物田间杂草识别及其它绿色植被分割中受光照变化影响较大的领域。

The solution of the separation of soybean seedling image and weeds from the soil background offers the premise and the foun- dation in terms of the soybean field are recognition. At present, in order to solve the problems like the influence of illumination on weed recognition, the poor adaptability to environment, 400 images concerning vegetation image and background color components under dif- ferent environments are statistically analyzed, obtaining that the identical pixel point consist of three color which are red （R）, green（G）, blue （B）,the G value of color components in green vegetation image is bigger than R and B values. However, the background is just the opposite. The experimental study indicates a very good effect which has been achieved regarding color characteristic division through G-R and G-B double thresholds value. This method compared with the 2G-R-B color characteristic is more effective in the vegetation and the soil background division, that a wider domain of its application to the field weed recognition, the green vegetation in complicated division etc.