基于图像处理的苦荞品种判别

为了建立不同品种苦荞种子的正确分类方法,以11个不同品种的苦荞种子为研究材料,使用扫描仪获取种子彩色图像,通过图像处理软件提取每个种子的特征变量28个,建立了一个包含11个品种440个苦荞种子,12 320个数据的矩阵。通过逐步法筛选有效特征变量,并利用有效变量构建判别模型。结果表明,筛选出的18个变量中,颜色变量为苦荞品种判别的主要变量。颜色变量和形态变量相结合,构建的贝叶斯判别模型,回判正确率达到了96.8%,交互验证正确率达到了94.7%。利用计算机图像处理技术和现代统计方法,能有效地对苦荞种子的颜色与形态特征进行精确量化和快速分析,可作为不同品种苦荞种子分类鉴别的一种客观,准确、有效的方法。

The Cultivar Discrimination of Tartary Buckwheat Based on Image Processing

The objective of this study is to establish a correct discriminant method for tartary buckwheat seeds of different varieties. The experiment with 11 different varieties of tartary buckwheat seeds as research materials, used a scanner to obtain seed color images, extracted 28 characteristic variables of each seed through image processing software, and then established a matrix which contains 12 320 data of 11 varieties, 440 tartary buckwheat seeds. The discriminant model was constructed by using the effective characteristics variables which has been screened through the stepwish method. The results showed that the color is the main variable in 18 variables selected for tartary buckwheat varieties identifying. The Bayes discriminant model was established with the combination of color and shape variables. Its accuracy of back substitution is 96.8%, and the accuracy of cross validation is 94.7%. Using computer image processing technology and modern statistical methods can extract and analyse buckwheat seed color and shape effectively. It is an objective, accurate and effective method for tartary buckwheat seeds classification.