关键词: 关键字：马铃薯干腐病, 时序高光谱, 时序关键点, 模糊聚类, 动态时间弯曲

Abstract: Abstract：The paper studies the diagnosis method of dry rot of the potato from incubation to symptom appearance by using time series hyperspectral to analyze and distract the disease features in the process of disease. Meanwhile, the pa per diagnoses the time series key point, namely the start point of symptom appearance by the means of DTW algorithm on the basis of the time series features. The paper uses threshold segmentation algorithm to distract ROI in the data preprocess, adopts the probability density algorithm to eliminate the abnormal data of disease spectrum. With the comparison of the spectrum and appearance of the disease, the study finds out the monotonicity characteristics in the spectrum of the dry rot of the potato, according to which the study distracts the spectrum feature by the means of KPCA. In the end, the study predicate time series key point by using fuzzy clustering model (FCM) based on the disease features, the right rate of which is only 66.7%. By contrast, the right rate of DTW on the basis of time series feature is as high as 94.4%. The study proves that the time series hyperspectral image diagnosis method based on DTW could diagnose dry rot of potato symptom appearance effectively, the result of which has important implication for reducing the spoilage rate of the potato in storage period.

Key words: Key words：Dry rot disease of potato, Time series hyperspectral, Time series key point, Fuzzy clustering, Dynamic Time Warping