【摘要】 为了提高玉米品种自动识别的可靠性,本文对表征品种的胚部特征参数进行了优化研究。采用区域生长法从玉米种子图像中分割出胚部区域,提取该区域的8个形状、6个颜色和6个纹理特征参数;定义了这些特征的类间和类内差异度计算公式,以便量化特征参数的有效性;结合改进的K-均值聚类算法,获得胚部形态的最优特征参数集。通过5个玉米品种各180粒的识别结果得知:玉米胚部特征参数在品种识别中作用显著,单纯基于胚部的优化特征参数集就可使其平均识别率达88%。本研究成果可为玉米品种自动识别开辟一条新思路。更多还原

【Abstract】 In the paper,a method for automatic identification corn varieties has been proposed. The method was K-means clustering algorithm combined degree of difference of characteristic based on corn kernel embryo morphology. It extracted the embryo region adopting region growing algorithm; extract characteristics of embryo region:eight shape features,six color features and six texture features. In order to select the most effective features of the embryo for identification of corn varieties,difference degrees of inter-class and intra-class of different feature for measure the effectiveness of features have been defined. K-means clustering algorithm with the characteristic difference degree has been used to find the optimal portray embryo morphology feature subset to recognize corn varieties.Five corn varieties were selected as the research object,180 kernels respectively. The average recognition rate was88% after researched by K-means algorithm with the feature subset.更多还原