摘要
在自动除草系统中优化杂草图像分割算法是降低识别误差的有效途径,为此提出了一种基于神经网络的分割算法。首先由训练样本统计出植被和背景在RGB颜色空间的分布概率,接着通过Bayes理论得出最优分割曲面训练BP神经网络,再通过BP神经网络将各种颜色分为植被和背景两类,并据此分割杂草图像。与其他三种杂草图像分割算法比较,新方法以颜色代替像素点为研究对象并据此构造最优分割曲面从而减小了分割误差并具备较好的泛化能力。
Optimizing weed image segmentation algorithm is being encouraged by concern over reducing identification errors in weed control system.A novel weed image segmentation algorithm based on neural network is proposed.The vegetable and background pixels probabilities in RGB color space are first got from training set.And then the BP neural network is trained by the optimization segmentation curve surface that is established with Bayes formula.All colors are divided into two parts with the BP neural network and each image is segmented with them.Compared with other three segmentation methods the results for the BP neural network show that the segmentation errors are reduced and the neural network is good generalization because of depending on colors instead of pixel and building optimization segmentation curve surface with them.
出处
《计算机工程与应用》
CSCD
北大核心
2009年第24期216-218,共3页
Computer Engineering and Applications
基金
广西工学院科学研究基金资助项目No0840107~~
