摘要
为了对蔬菜病害进行自动检测和分级,克服传统的评估方法检测成本高、划分等级随意性比较大的缺陷,探讨了基于蔬菜图像的病害等级自动分类,设计了预处理技术提取图片特征.具体方法是对叶片图像进行分割并转变为分布概率的灰度矢量.在此基础上利用数据挖掘技术建立决策树模型,利用该模型对未分级叶片的级别进行预测.还引入了元挖掘算法Bagging,与决策树模型综合使用,进一步提高了预测效果.实验表明这一蔬菜病害等级自动分类方法准确度较高,具有较好的实用价值.
Traditional method for classifying the grade of vegetable （cucumber） disease is mainly based on the decision of human observation. The disadvantage of this method is that it is commercially costly and unstable, different observer may have different back knowledge and different decision. This paper discusses the automatic classification of the vegetable disease grade based on vegetable images. The paper designs preprocessing method to extract features from these images, transform the images of cucumber leaves to data vector. Then, according to the data whose disease grade has been identified by experts, decision tree is built to get a relative accurate prediction model. In addition, a meta mining method, bagging, is introduced to be combined with decision tree to achieve even better prediction accuracy. Experimental results show that the scheme proposed is successful in accurately predicting the vegetable disease grade.
出处
《北京师范大学学报（自然科学版）》
CAS
CSCD
北大核心
2006年第6期592-596,共5页
Journal of Beijing Normal University(Natural Science)
基金
国家自然科学(青年)基金资助项目(1000100660273015)
关键词
蔬菜病害等级分类
图像处理
数据挖掘
决策树
元挖掘
classification of the vegetable disease grade
image processing
data mining
decision tree
meta mining
