【摘要】 利用MS3100多光谱相机采集了受稻瘟病侵染的秧苗多光谱图像,通过图像分割、复原和分析,得到了在近红外、红光和绿光波段稻苗植株样本图像的灰度均值。将3个波段的图像灰度值作为特征参量,并采用支持向量机技术建立稻瘟病的抗病性分级检测模型。结果表明,模型具有较高的分类精度。抗性样本和感病样本的分类精度达到100%,抗性样本和中等感病样本的分类精度为96.8%。本研究为水稻品种抗病性鉴定调查提供了一种新的方法,同时也为稻瘟病早期检测提供了基础。更多还原

【Abstract】 Rice blast resistance identification based on multi-spectral computer vision technology was studied.Multi-spectral camera(MS3100) was used to capture rice seedlings infected by blast sample images,and then image gray values of near infrared(NIR)、red(R) and green(G) were measured by image processing and analysis.The support vector machine method was adopted to build rice blast resistance identification model,and image gray values of three bands (NIR、R and G) as parameters were put into the model.The results show that the classing precision can achieve 100% of healthy samples and sensing disease samples,and 96.8% of healthy samples and middling sensing disease samples.The results provide a new method for rice disease resistance identification and made a preparation for early detection on rice blast.更多还原