基于计算机视觉的三七主根质量的分级方法

发布时间：2018-10-25 06:22

【摘要】：选取干燥后的三七主根样本110个,运用计算机视觉技术获取三七主根样本图像,对图像进行灰度化、二值化以及形态学运算,提取长、宽、投影面积等特征值。结果表明,三七主根的形状可分为锥形和瘤形,分别对2种主根建立投影面积和质量的关系预测模型,三七主根的质量和投影面积呈线性相关,锥形三七主根与瘤形三七主根投影面积和质量预测模型的决定系数R2分别为0.984 9和0.986 6。采用十折交叉验证法对质量预测模型进行验证,锥形三七主根质量误差均值0.334 8 g;瘤形三七主根质量误差均值0.494 9 g。
[Abstract]:In this paper, 110 samples of main root of Panax notoginseng after drying are selected, and the image of main root of Panax notoginseng is obtained by computer vision technology. The image is grayscale, binarized and morphologically calculated, and the characteristic values such as length, width and projected area are extracted. The results showed that the shape of the main root of Panax notoginseng could be divided into conical and nodular shapes. The relationship between projection area and mass was established for the two main roots, and there was a linear correlation between the mass and projection area of the main root of Panax notoginseng. The determination coefficient R2 of projection area and mass prediction model of main root of Panax notoginseng and Notoginseng notoginseng are 0.984 9 and 0.986 6, respectively. The method of 10 fold cross validation was used to verify the quality prediction model. The mean value of mass error of the main root of Panax notoginseng was 0.334 8 g, and that of the main root of notoginseng was 0.494 9 g.
【作者单位】：昆明理工大学现代农业工程学院;昆明理工大学工程训练中心;
【基金】：国家自然科学基金项目(11226220) 云南省科学技术厅项目(2010ZC028)
【分类号】：R284.1;TP391.41

【相似文献】