中文摘要：

为了实现准确无损的小麦品种检测，提出了将生物超微弱发光信息与模式识别技术相结合的小麦品种聚类分析方法。通过测量不同品种小麦发射的生物光子，分别使用均值、方差和自相关值作为特征向量，然后使用K—means和ISODATA算法对这些特征向量进行处理，实现小麦品种的识别。实验结果表明，使用提出的方法可以实现小麦品种的无损检测，同时也表明单独使用均值和方差这两维特征能使分类效果达到更好。所提出的方法可用于小麦品种的早期辅助决策，指导粮库科学有效地开展收购储藏等工作。

英文摘要：

In order to achieve accurate nondestructive detection of wheat varieties, we proposed a clustering analysis method b.y combining the biological ultraweak luminescence information and pattern recognition technique. The biophoton emissions of different wheat varieties were measured with mean values, variances and autoeorrelation as feature vector, these feature vectors were processed by K-means and ISODOTA algorithm methods to identify the wheat varieties. The results showed that the method could realize the nondestructive detection of wheat varieties, and separately use of the two dimensional feature of mean and variance could achieve better classification effect. The proposed method could he used in early auxiliary decision for wheat varieties, and for grain storage guidance.