基于Logistic、IBk以及Randomcommittee方法的条锈病潜育期小麦冠层光谱的定性识别
Qualitative identification of canopy spectra in wheat stripe rust based on Logistic, IBk and Randomcommittee methods

为寻求在小麦条锈病潜育期能探知和监测病害的简单便捷方法，通过人工接种不同品种小麦诱发条锈病，在小麦条锈病菌尚处于潜育期时，采集小麦冠层光谱数据，并利用双重Real-timePCR分子生物学技术检测条锈病菌潜育菌量，基于Logistic、IBK以及Randomcommittee三种方法，在不同建模比、不同参数变换下建立可识别潜育期小麦条锈病的数学模型。结果表明，在全波段范围内（325~1 075 nm），3种方法所建模型模拟识别潜育期小麦条锈病是可行的，但识别效果有一定差异，基于Logistic、IBK以及Randomcommittee方法所建模型的平均准确率分别为83.95%~84.51%、87.72%~88.98%、93.19%~93.46%。因此，基于Randomcommittee方法所建模型的识别准确率最高，效果最好，更适合小麦条锈病潜育期的定性识别。
To explore the rapid diagnosis method of wheat stripe rust during the latent period, different varieties of wheat were artificially inoculated by the Puccinia striiformis f. sp. tritici(Pst). The canopy hyperspectral data was collected in the latent period, and the amount of Pst was also obtained by using the duplex Real-time PCR. Based on the three methods of Logistic, IBk and Randomcommittee, the hyperspectral remote sensing mathematical models were established to recognize the wheat stripe rust during the latent period with different modeling ratio and different modeling parameters. The results showed that within the 325-1 075 nm waveband, the mathematical models based on the methods of Logistic, IBK and Randomcommittee to discriminate wheat stripe rust in the latent period was feasible. But a certain difference in the recognition effectiveness was also found. The average recognition accuracy of the Logistic, IBK and Randomcommittee methods were 83.95%-84.51%, 87.72%- 88.98%, 93.19%-93.46%, respectively. The results indicated that the mathematical model based on Randomcommitteemethod weremore suitable for qualitative identification of wheat stripe rust during the latent period.