摘要
为了实现作物病害的计算机识别,提出了一种基于双编码遗传特征选择的支持向量机和病害图像多特征参数识别病害的方法。通过病害图像增强处理,采用基于HIS颜色空间的H分量与大津法(Otsu)结合对病斑图像自动分割,自动提取病斑的特征参数;运用双编码遗传算法优化病斑特征子集,并对其赋予权重,底层构建一对一投票策略的支持向量机分类识别作物病害的方法和途径。应用该方法对烟草病害中多种容易混淆的病害进行实验,结果表明:该方法与没有采用遗传算法的支持向量机相比,在同等条件下,特征向量减少了38%,而正确率却提高了6.29%,具有一定的有效性和实用价值。
To realize computer identification of crop diseases, this paper puts forward a method based on the dual coding genetic characteristics of support vector machine and multi-feature parameters of the disease image to identify disease. The article attempts the ways and means by the disease image enhancement, using Otsu method （Otsu） combined lesion image automatic segmentation based on the H component of the HIS color space, automatic extraction of the lesion char-acteristic parameters, using dual coding genetic algorithm to optimize the lesion subset of features, and its assign weights, the underlying construct a one-to-one voting strategies support vector machine classifier to identify crop diseases. Appli-cation of the method to the experimental results show that ： the method compared with support vector machines of no using genetic algorithms feature vectors, under the same conditions, a 38% reduction in confusing the disease on a variety of tobacco diseases, and correct rate improve6.29%, and the effectiveness and practical value.
出处
《农机化研究》
北大核心
2014年第1期60-64,共5页
Journal of Agricultural Mechanization Research
基金
云南省自然科学研究基金重点项目(2012Z116)
关键词
自动识别
支持向量机
病害图像
特征向量
双编码遗传算法
automatic identification
support vector machine （SVM）
image diseases
feature vector
double coding ge-netic algorithm
