基于压缩感知理论的苹果病害识别方法  

Apple Disease Recognition Based on Compressive Sensing

作　　者：霍迎秋[1,2] 唐晶磊[2] 尹秀珍[2] 方勇[2] 

机构地区：[1]西北农林科技大学机械与电子工程学院,陕西杨凌712100 [2]西北农林科技大学信息工程学院,陕西杨凌712100

出　　处：《农业机械学报》2013年第10期227-232,共6页Transactions of the Chinese Society for Agricultural Machinery

基　　金：国家自然科学基金资助项目(61271280;61001100);陕西省自然科学基金资助项目(2010K06-15)

摘　　要：为实现自然场景下低分辨率苹果果实病害的智能识别,提出了一种基于压缩感知理论的苹果病害识别方法。以轮纹病、炭疽病和新轮纹病3种常见的苹果果实病害为研究对象,提取病斑的8个纹理特征参数组成训练特征矩阵。利用压缩感知理论,求解待测样本特征向量在特征矩阵上的稀疏表示系数向量,通过对系数向量的分析实现待测样本的分类。设计灰度关联分析和支持向量机识别模型与本文方法进行识别效果对比,平均正确识别率分别为86.67%、90%和90%。实验结果表明,基于压缩感知理论的识别方法能够对苹果病害进行有效识别。To intelligently recognize apple fruit diseases from low-resolution images taken in natural environment, a method based on compressive sensing was proposed. Three kinds of apple fruit diseases （apple ring rot, apple anthracnose and new apple ring rot） were investigated. Eight texture feature values were extracted to construct the training eigenmatrix. Then compressive sensing was used to approximate the sparse coefficient vector which was the sparse representation of the sample eigenvector on the training eigenmatrix. Thus the test sample was classified by analyzing the coefficients vector. Both the gray relation analysis and the support vector machine recognition models were constructed to compare with the proposed method. The recognition rates of three models were 86.67% , 90% and 90% , respectively. The experimental results showed that the recognition method based on compressive sensing could effectively recognize these three kinds of apple fruit diseases.

关 键 词：苹果病害 压缩感知 特征矩阵 稀疏表示 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术] S436.611.1[自动化与计算机技术—计算机科学与技术]