基于无人机数字图像与高光谱数据融合的小麦全蚀病等级的快速分类技术

小麦全蚀病是检疫性的土传病害,对小麦生产危害极大,对其发生的监测是治理的根本。遥感技术可实时、宏观地监测病害发生发展,尤其是将光谱信息与高分辨率数字图像进行融合,可直观、精准地对病害识别和分类。本文基于计算机视觉技术,通过光谱数据与高分辨率数字图像结合的方法,对小麦全蚀病等级进行快速分类。首先,通过ASD非成像光谱仪获取小麦全蚀病的光谱信息,提取全蚀病特征光谱,建立光谱比。其次,利用无人机获取的实时田间数码图像,对其颜色特征进行重量化。最后,利用基于支持向量机的决策树分类对图像视场中的不同全蚀病等级进行分类。结果表明,4个全蚀病等级的分类精度均大于86%(Kappa0.81),平均运算时间小于30s。通过与实地调查的小麦全蚀病的白穗率等级做比对,验证分类结果的准确性,结果表明该方法基本可以实现对小麦全蚀病等级的实时监测。

Fast multi-classification of wheat take-all levels based on the fusion of unmanned aerial vehicle digital images and spectral data

Wheat take-all will lead to a disaster in wheat production without timely monitoring and management. Traditional remote sensing approaches in wheat take-all have failed to fast and accurately recognize the multi-level disease conditions due to relatively coarse spatial resolution and the experience-based features selection. This study developed a method to achieve the fast multi-classification of wheat take-all based on the computer vision and the data fusion technology. Firstly, ASD HandHeld sensor was used to extract the spectral feature ratio. Then the color model was established to quantify the UAV aerial photo. Finally, the wheat take-all were classified using the decision tree which based on the support vector machine (SVM). The results showed that an overall accuracy was greater than 86% (Kappa > 0.81) for classifying all of take-all levels, and computation rate was less than 30 seconds, which is meaningful for automatic real-time monitoring of take-all conditions.