基于机器视觉的三七主根形状分类方法研究

A Research on the Shape Classification Method of Panax Notoginseng Taproot Based on Machine Vision

从计算机视觉的角度对锥形和瘤形的三七主根进行形状识别。采集若干个三七主根的正投影图像,研究三七主根形状与其正投影图像几何特征的关系。将三七主根正投影图像的圆形度、矩形度、和轴长比三个参量同时作为特征值,借助matlab神经网络工具箱,融合定性变量数值化的方法,构建形状判别函数,实现更为准确地对非常规形状的三七主根形状进行判别。实测结果表明,对于非常规形状的三七主根,本文提出的分类方法优于传统区分方法。

Shape recognition of cone-shaped and nodular Panax notoginseng taproots is done from the perspective of computer vision.Several orthographic images of the main root of Panax notoginseng were collected,and the relationship between the shape of the main root of Panax notoginseng and the geometric characteristics of the orthographic image was studied in our research.The three parameters of circularity,rectangularity,and axial length ratio of the orthographic projection image of the main root of Panax notoginseng are taken as eigenvalues at the same time.With the help of the matlab neural network toolbox,the method of quantifying qualitative variables is combined to construct a shape discriminant function to achieve more accurately discrimination of the shape of unconventional shape Panax notoginsenge.The experimental results show that the discriminant method proposed in this paper is better than the traditional discrimination method for the unconventional shape of the main root of Panax notoginseng.