基于图像的小麦白粉病形状特征参数优化与提取

发布时间：2018-12-16 13:32

【摘要】：精准识别作物病害的前提是准确地提取病害特征值,针对现有研究主要侧重于增加特征参量个数来提高病害识别率。以小麦白粉病为例,充分考虑形状描述子的平移及旋转不变性,在原有矩形度、伸长度等特征值的基础上重新构造病斑的圆形度、曲率等特征描述参数作为特征值,进行单个病斑参数的提取试验,并对获得的参数进行对比分析,以选出描述病斑形状的最优参数及其数值范围。结果表明,选用的特征参数值中圆形度C2、矩形度、伸长度及曲率具有较小的分散性,能精确有效地表征病斑的形状特征,显著提高病斑描述的精确性;并求取相关特征参数值的大致取值范围,增加人们对白粉病的认识,为小麦白粉病的自动识别提供数据基础。
[Abstract]:The premise of accurate identification of crop diseases is to extract the disease eigenvalues accurately. The existing researches mainly focus on increasing the number of feature parameters to improve the disease recognition rate. Taking wheat powdery mildew as an example, the translation and rotation invariance of the shape descriptor are fully considered. Based on the original eigenvalues such as rectangle degree and elongation degree, the roundness and curvature of the disease spot are reconstructed as the eigenvalues. In order to select the optimal parameter and its numerical range for describing the shape of the disease spot, the extraction experiment of single disease spot parameter was carried out, and the obtained parameters were compared and analyzed. The results show that the roundness C _ 2, rectangle, elongation and curvature of the selected characteristic parameters have small dispersity, which can accurately and effectively characterize the shape characteristics of the disease spot and improve the accuracy of the disease spot description. The range of relative characteristic parameter values is obtained to increase people's understanding of powdery mildew and to provide data basis for automatic identification of wheat powdery mildew.
【作者单位】：郑州轻工业学院电气信息工程学院;
【基金】：国家自然科学基金(编号:61501407) 河南省科技厅科技攻关项目(编号:162102110118) 河南省高等学校重点科研项目(编号:15A413006) 郑州轻工业学院研究生科技创新基金(编号:2015028)
【分类号】：S435.121.46;TP391.41

【相似文献】