文章摘要
龙满生,刘清,曾小荟.基于颜色特征的油茶叶片病斑分割研究[J].井冈山大学自然版,2017,(2):47-54
基于颜色特征的油茶叶片病斑分割研究
SEGMENTATION OF CAMELLIA LEAF LESION BASED ON COLOR FEATURES
投稿时间:2016-10-08  修订日期:2017-02-20
DOI:10.3969/j.issn.1674-8085.2017.02.010
中文关键词: 图像分割  油茶  叶片病斑  颜色特征
英文关键词: image segmentation  camellia oleifera  leaf lesion  color feature
基金项目:流域生态与地理环境监测国家测绘地理信息局重点实验室开放基金课题(WE2015002);江西省教育厅科技落地计划项目(KJLD13066)。
作者单位E-mail
龙满生 井冈山大学电子与信息工程学院, 江西, 吉安 343009
流域生态与地理环境监测国家测绘地理信息局重点实验室, 江西, 吉安 343009 
longmansheng@126.com 
刘清 井冈山大学电子与信息工程学院, 江西, 吉安 343009
流域生态与地理环境监测国家测绘地理信息局重点实验室, 江西, 吉安 343009 
 
曾小荟 井冈山大学电子与信息工程学院, 江西, 吉安 343009
流域生态与地理环境监测国家测绘地理信息局重点实验室, 江西, 吉安 343009 
 
摘要点击次数: 2236
全文下载次数: 2537
中文摘要:
      油茶叶片病斑分割是提高油茶病害图像识别准确率的前提。根据油茶炭疽病、软腐病和煤污病等主要病害的症状特征,提出了基于超绿特征的油茶叶片病斑检测方法。对比分析了HSI、YCbCr和Lab颜色模型的各个颜色分量以及超绿特征对油茶叶片病斑分割性能的影响。对于炭疽病和软腐病,超绿特征的单阈值分割性能最好,色度a的分割性能较好,色差Cb和色差Cr的分割性能稍差,色调H的分割性能最差,其Ⅰ型错误率很高。对于煤污病,亮度L的分割效果最好,亮度Y的分割效果次之。试验结果表明,超绿特征对于三种典型油茶病害图像的综合分割性能较高,具有较低的Ⅰ型错误率和Ⅱ型错误率,其平均检测精度达到81.69%以上,可以降低因叶柄和叶脉影响而产生的分割错误。
英文摘要:
      Segmentation of Camellia leaf lesion is the premise to improve the accuracy of Camellia disease image recognition. According to the symptom characteristics of Camellia anthracnose, agaricodochium and sooty mold, a leaf lesion detection based on excess green feature (2G-R-B) was put forward. A comparative analysis of segmentation performance was carried out among the excess green feature, and each color component of HSI, YCbCr and Lab color spaces. For the anthracnose and agaricodochium, the segmentation performance of excess green feature with a single threshold is the best, chroma A is the second, chromatic aberration Cb and Cr are slightly worse, and hue H is the worst with very high type I error rate. For the sooty mold, the segmentation performance of lightness L with a single threshold is the best, and brightness Y is the second. The experimental results show that excess green feature can obtain high overall segmentation performance for typical diseases of Camellia leaf with low type I and type Ⅱ error rates, which can reduce the segmentation error due to the impact of petiole and leaf vein, and its segmentation accuracy reaches over 81.69%.
查看全文   查看/发表评论  下载PDF阅读器
关闭