基于高光谱成像技术的水稻叶瘟病斑分割方法及其光谱特性分析.docVIP

  基于高光谱成像技术的水稻叶瘟病斑分割 方法及其光谱特性分析# 郑志雄1，齐龙2，马旭2，谢俊锋1** 5 10 15 20 25 30 （1. 华南农业大学工程学院，广州 510642； 2. 华南农业大学南方农业机械与装备关键技术教育部重点实验室，广州 510642） 摘要：利用 HyperSIS 高光谱成像系统采集了受稻瘟病侵染的水稻叶片的高光谱图像, 通过 主成分分析方法得到了利于病斑识别的第 2 主成分图像。在此基础上，采用密度分割法对其 进行图像分割，有效地识别了病斑，并分析了叶瘟病斑区域与正常叶片部分的光谱特征差异。 结果表明：在采集的 60 个受稻瘟病侵染的水稻叶片样本中，其中 54 幅图像中的病斑部分可 被准确分割，分割率为 90%；绿光波段（530~580nm）、红光波段（600~760nm）和近红外 波段（720~988nm）处水稻叶瘟病部位与正常叶片相比，光谱反射率分别呈现出下降、上升 和下降的趋势，且在 600~690nm 波长区域内差异明显。本研究为稻瘟病田间检测提供了基 础。 关键词：水稻叶瘟病；高光谱成像；主成分分析；图像分割；光谱分析 Segment Method and Spectral Features Analysis of Rice Leaf Blast Using Hyperspectral Imaging Technology Zheng Zhixiong1, Qi Long2, Ma Xu2, Xie Junfeng1 (1. College of Engineering , South China Agricultural University, GuangZhou 501642; 2. Key Laboratory of Key Technology on Agricultural Machine and Equipment, Ministry of Education , South China Agricultural University, GuangZhou 501642) Abstract: The hyperspectral image of leaves infected by leaf blast was captured with HyperSIS hyperspectral system, and the second principal component image (PC2) was obtained to identify blast by using principal component analysis (PCA). On the basis, the leaf blast area was efficiently identified using density slice, and spectral features of leaf blast areas and normal areas were analyzed. The result showed that in the acquisition of 60 samples infected by rice leaf blast, 54 images with blast can be accurately segment and the overall segmentation rate was 90%; compared to the healthy leaves, the reflectance rate at green (530~580nm), red (600~760nm) and near infrared (720~988) bands was down, up and down, and the differences of reflectance rate were significant in the regions of 600~690nm. This research provided some advice for field detection on rice blast. Keywords: rice leaf blast; hyperspectral imaging; principal component analysis; image segmentation; spectral feature analysis 35 0 引言 稻瘟病是我国南北稻作区危害最严重的水稻病害之一，与纹枯病、白叶枯病并称为水稻 三大病害[1]。目前稻瘟病的识别主要还是人工通过图片对比或根据文字描述来完成，然而这 些识别方法主观性太强，对工人专业素质要求较