摘要
【目的】探索基于形状特征和支持向量机(SVM)的茶叶病害识别方法,为茶叶病害的智能准确识别提供技术支撑。【方法】采集贵州铜仁茶区茶炭疽病、茶饼病、茶白星病的病斑图像,使用MATLAB提取并计算3种病害的病斑面积、周长、外接矩形和外接椭圆面积、复杂性、伸长度、矩形度、圆度、面积凹凸比8种形状特征值,分别建立基于单一形状特征和复杂性、伸长度、矩形度、圆度及面积凹凸比5种组合形状特征、4种不同核函数的SVM,对3种茶叶病害进行分类识别,比较其正确识别率,筛选最优识别算法。【结果】基于单一特征的SVM对3种茶叶病害的识别率低于组合形状特征下的正确识别率;基于组合形状特征的SVM中,线性核函数、多项式核函数、径向基核函数、Sigmoid核函数的SVM总识别率分别为90.00%、88.00%、83.33%、86.05%。【结论】基于复杂性、伸长度、矩形度、圆度及面积凹凸比5种组合形状特征,采用线性核函数的SVM对茶炭疽病、茶饼病、茶白星病的分类识别效果较优。
【Objective】In order to provide the technological support for intelligent and accurate recognition of tea disease,the recognition method of tea disease is explored based on shape features and support vector machine(SVM).【Method】The disease spot images of tea anthracnose,tea gall and white scab disease in Tongren tea-growing area of Guizhou are collected.MATLAB is used to extract and calculate eight shape eigenvalues of disease spot area,perimeter,bounding rectangle and circumscribed ellipse area,complexity,elongation,rectangular degree,roundness and area concave-convex ratio in three diseases.SVMs are respectively established based on single shape feature,and four different kernel functions of five combined shape features including complexity,elongation,rectangular degree,roundness and area concave-convex ratio to classify and identify the three tea diseases,and then the correct recognition rate is compared and the optimum recognition algorithm is screened.【Result】The recognition rate of SVM based on single feature toward three tea diseases is lower than the correct recognition rate under the combined shape features.Based on the SVM of combined shape features,the total SVM recognition rate of linear kernel function,polynomial kernel function,radial basis kernel function and Sigmoid kernel function is 90.00%,88.00%,83.33%and 86.05%respectively.【Conclusion】SVM with linear kernel function based on the five combined shape features of complexity,elongation,rectangular degree,roundness and area concave-convex ratio has the better classification and recognition effect on tea anthracnose,tea gall and white scab disease.
作者
陈荣
李旺
周文玉
CHEN Rong;LI Wang;ZHOU Wenyu(School of Data Science,Tongren University,Tongren,Guizhou 554300;Tongren Weituo Network Technology Co.Ltd.,Tongren,Guizhou 554300,China)
出处
《贵州农业科学》
CAS
2021年第4期53-59,共7页
Guizhou Agricultural Sciences
基金
贵州省教育厅自然科学研究项目“基于电子舌技术的石阡苔茶品质检测研究”[黔教合KY字(2019)181]
贵州省大学生创新创业训练计划项目“基于BP神经网络的茶叶病害识别研究”(20195201739)
铜仁市科技计划项目“计算机视觉在梵净山区茶叶品种鉴别中的应用研究”[铜市科研(2019)96]。
关键词
茶叶
病害识别
形状特征
支持向量机
图像处理
机器学习
tea
disease recognition
shape feature
support vector machine
image processing
machine learning
