基于颜色特征和属性约简的黄瓜病害识别方法

谢泽奇,张会敏,张善文,张云龙

(郑州大学西亚斯国际学院，河南郑州451150)

摘要：为了减少黄瓜叶部病害给农业生产带来的损失，提高病害的识别率和精度，提出了一种基于颜色特征和属性约简算法的黄瓜病害叶片分割与识别方法。该方法首先利用最大类间方差(Otsu)阈值法对黄瓜病害叶片图像进行病斑分割；其次提取病斑图像的36个分类特征，再利用基于区分矩阵的属性约简算法进行特征选择；最后利用最近邻分类器进行病害识别。该方法在5种常见黄瓜病害叶片图像数据库上进行了病害识别试验，结果表明，识别率高达 94.8%。说明，该方法对作物病害叶片图像识别是有效可行的。

关键词：颜色特征；属性约简；病斑分割；病害识别

doi:10.3969/j.issn.1000-4440.2015.03.010

收稿日期：2014-11-19

基金项目：国家自然科学基金项目(61272333)；河南省科技厅科技攻关项目(142102310518、142400410853、142300410309)；河南省教育厅科学技术研究重点项目(14B520064、15A520100)；郑州大学西亚斯国际学院校级科研项目(2014KYYB23)

作者简介：谢泽奇(1981-),男,河南镇平人,硕士，讲师，研究方向为计算机应用、图像处理。(E-mail)xzq0413@163.com

通讯作者：张会敏,(E-mail)zhm0413@163.com

中图分类号：TP391

文献标识码：A

文章编号：1000-4440(2015)03-0526-05

Abstract：To reduce the loss caused by cucumber leaf disease, and improve disease recognition rate and accuracy, a leaf image segmentation and disease recognition method for cucumber was proposed based on color feature and attribute reduction algorithm. Firstly, the Otsu algorithm was applied to segment the cucumber diseased leaf image. Secondly, 36 diagnostic characters of disease lesion were extracted, and selected by using attribute reduction algorithm based on discernibility matrix. Finally, the crop diseases were recognized using the nearest neighbor classifier. As an effective and feasible approach for crop disease recognition, this method could recognize as high as 94.8% of five cucumber diseases.

Cucumber disease recognition based on color feature and attribute reduction

XIE Ze-qi,ZHANG Hui-min,ZHANG Shan-wen,ZHANG Yun-long

(SIASInternationalUniversity,ZhengzhouUniversity,Zhengzhou451150,China)

Key words：color feature;attribute reduction;image segmentation;disease recognition

黄瓜病害对其产量和质量带来了巨大损失，如何利用计算机图像处理技术在黄瓜种植期间实现病害及时检测识别和自动防治在农业物联网技术研究领域具有重要意义。由于实际病害叶片形状、纹理和颜色的多样性，使得叶片图像中提取的病害分类特征也各种各样，因此，需要进行特征选择。……