Key points are not available for this paper at this time.
Abstract Statistical methods like principal component analysis and cluster analysis are not new in identification and classification for biological features. However, the success of utilizing these two methods in discriminating late blight infected tomatoes (caused by Phytophthora infestans ) from healthy ones has not yet been reported. This paper demonstrates the capability of using principal component analysis and cluster analysis for identification and discrimination of spectral characteristics of late blight infections on tomatoes. Our results show that the first principal component is related to the spectral properties of healthy tomatoes, and the second principal component is related to the spectral properties of infected tomatoes. Cluster analysis shows that a reasonable discrimination is obtained when the centroid distance of clusters is above 0.5. The consistent results from both principal components analysis and cluster analysis indicate that late blight infection on tomatoes can be successfully detected with remote sensing when the infection severity reaches middle to late stages. Moreover, spectral ratio analysis provides us with the way to identify the sensitive spectral wavelengths where distinguishable reflectance values can be observed for unique biological features. Understanding the light responses to unique biological features may increase discrimination accuracy by reducing the impact of soil background on spectral measurements, and utilizing the most sensitive wavelengths for discriminating between healthy and diseased tomatoes.
Zhang et al. (Tue,) studied this question.