摘要
密集烤房内烤烟烘烤阶段的自动识别是建立智能化烟叶烘烤系统的重要环节.为了有效地识别烤房内烤烟的烘烤阶段,该文提出了一种基于图像特征和GA-SVM(Genetic Algorithm-Support Vector Machine)相结合的方法.该方法将机器视觉系统提取的烤烟图像特征作为SVM的输入参数,通过GA全局搜索特性选取出模型的最优特征子集,最后通过多分类SVM实现对烘烤阶段的识别,同时验证了选取特征的有效性.仿真结果表明:从9个原始特征中筛选出5个图像特征,总体识别精度从93.7%提高到96.5%,能有效地识别烤烟的烘烤阶段,具有良好的在线应用前景.
The recognition of tobacco flue-curing phases in bulk curing barn is an important part of building intelligent tobacco curing system. In order to recognize the curing phase effectively, a combination method based on image features and GA-SVM algorithm has been proposed. The proposed method uses machine vision system to extract the tobacco images features which are input to SVM. The optimal feature subset has been selected thanks to the global search ability of GA. The curing phase has been recognized by multi-classifier SVM and the effectiveness been verified either. Simulation results demonstrate that 5 image features are selected from original 9 features, which makes the overall recognition accuracy increased from 93. 7% to 96. 5%. The proposed method can effectively recognize tobacco curing phases and has good prospects for online applications.
出处
《西南师范大学学报（自然科学版）》
CAS
北大核心
2016年第9期100-106,共7页
Journal of Southwest China Normal University(Natural Science Edition)
基金
重庆市教育科学"十二五"规划项目(2013-ZJ-081)
关键词
图像特征
遗传算法
支持向量机
烤烟
特征优选
image features
Genetic Algorithm
Support Vector Machine
tobacco
features selection
