Abstract: Grade classification of seed cotton is a major issue that has a significant impact on the agricultural economy. According to the characteristics such as impurities, yellowness and brightness extracted from images of seed cotton, fuzzy pattern recognition was used to improve the classification of cotton grade. A classification model of seed cotton was constructed based on the fuzzy nearness. Fuzzy mathematics was combined with artificial neural network to build up a well improved model and algorithm. Statistical distribution was used to calculate and select the model parameter method. Eventually, the numbers of impurities of different sizes were worked out by using the Euler's numbers of the image. Based on the method of selecting model parameters, the proposed algorithm could be optimized step by step. After full learning, seed cotton classification accuracy rate reached 92%. The experimental results show that the presented algorithm satisfies the actual application needs.

Key words: seed cotton, grade, fuzzy mathematics, pattern recognition, neural network

摘要： 籽棉品级分类问题是对农业经济有着重要影响的一个问题。在对籽棉图像黄度、亮度和杂质等特征提取分析基础上，基于模糊模式识别，运用模糊贴近度，构建籽棉品级分级模型，利用统计分布计算得出模型参数选取方法。利用图像欧拉数求得了不同大小杂质数量的近似值，运用神经网络对模型进行有效求解，通过调整模型参数使籽棉品级分级精度不断提高，分级模型在充分学习后，籽棉品级分级准确率达到92%，满足了实际应用的需要。

关键词: 籽棉, 品级, 模糊数学, 模式识别, 神经网络