[Objective]The aim was to realize fast and automatic dection of rice seed vigor.[Method]Hyperspectral imaging covering the spec-tral range of 874-1734 nm was applied to detect conventional rice seed vigor.Two rice seed cultivars ( named Yongxian69 and Zhongzao 39) were used for analysis.Spectral information of untreated rice seeds, rice seeds accelerated aging for 48 hours and 72 hours was extracted from hyperspectral images.Wavelet transform was applied to eliminate obvious random noises of pixel-wise spectra before spectra extraction.Sup-port vector machine (SVM) was applied to build discriminant models using full spectra.The SVM models of the two rice cultivars showed ac-ceptable results.[ Result] The untreated seeds and treated seeds could be accurately discriminated , while seeds accelerated aging for 48 hours and 72 hours could not be discriminated.The results matched with the vigor test by traditional methods, and the accelerated aging showed differences on different seed cultivars.[Conclusion]The hyperspectral imaging combined with chemometric methods could be used for fast, automatic and non-invasive detection of rice seed vigor.%[目的]实现常规水稻种子活力的快速自动化检测.[方法]采用高光谱成像技术(波长范围874~1734 nm),提取甬籼69和中早392种常规水稻种子未老化、老化48 h和老化72 h的光谱反射率,在提取样本光谱时采用小波变换(WT)剔除像素点光谱噪声部分,并基于全波段光谱建立了支持向量机(SVM)判别分析模型.[结果]未老化种子与老化种子可以准确识别,而老化48 h种子与老化72 h种子之间无法准确识别,与基于种子活力参数的测量结果相符,且不同水稻品种对老化的反应存在差异.[结论]高光谱成像技术结合化学计量学方法用于种子活力的快速自动化无损检测是可行的.
展开▼
