基于计算机视觉的红枣形状识别方法研究

Research on red dates shape recognition method based on computer vision

以红枣为研究对象,通过图像处理技术获得红枣的边界轮廓。用极半径函数来表示红枣的边界轮廓,将极半径函数用傅里叶级数展开,并使用傅里叶级数前15项系数来描述红枣的形状特征。然后分别使用欧氏距离法和不规则度判别法对红枣的形状进行了分类实验研究,实验结果表明:2种方法对正常枣的分类准确率都比较高;对于畸形枣,欧氏距离法的识别率仅为35%,而不规则度判别法的识别率可达90%。

Based on image processing technology,the contour profile of Chinese red dates is extracted.Polar radius function is used to describe the contour profile,and polar radius function is expanded by Fourier series and the top 15 terms of the series are selected to describe the shape feature of the contour profile.Euclidean distance and irregularity degree discrimination method are applied to recognize the shape of red dates respectively,and experimental results indicate that the two methods have high accuracy of recognition on normal red dates,on malformation ones,the recognition rate of Euclidean distance is only 35 % and that of irregularity degree method can reach 90 %.