边界点矩特征傅里叶描述的马铃薯薯形研究

Study on Shape of Potatoes Based on Dividing Point's Proper of Matrix's Fourier Descriptor

马铃薯的形状是马铃薯分级的重要指标之一。针对适用于实时图像处理的目标识别系统,本文提出了一种结合矩特征和傅立叶描述子的形状识别新方法。该方法以质心为中心将物体划分为多个扇形区域,计算各扇形区域的矩特征值获得表示物体形状的矩特征序列,再通过离散傅立叶变换得到具有平移、旋转以及比例不变性的归一化矩特征傅立叶描述子,采用相似度计算进行形状分类。实验结果表明,该方法对目标形状的平移、旋转和比例变换具有不变性,能准确地将马铃薯的形状分为椭圆、圆和畸形三类,准确率分别为90%,93.3%,100%,识别率较高,具有良好的应用前景。

Shape of potato is one of the important index of potato＇s classification. Amming at the object recognising system of real time figure processing, this article came up with a new shape recognising method which combined proper of matrix and Fourier descriptor. This method took center of mass as focus and divided object into many fan regions, computed the proper value of matrix of every fan region and got the proper of matrix series which can show objecCs shape. Then the normalization proper of matrix Fourier descriptor which was invariable thouth translation, rotation and proportion invariant was gotten via discrete Fourier Transformation. And shapes were classified thougth simi- larity calculation. The experimental result demonstrated that this method can make translation, rotation and proportion transformation un- changeable. And it can define shape of potatos as round, elliptical and anamorphotic accurately, accuracy rate are 90%, 93.3%, 100% respectively. Which has high recognition rate and a promising application prospect.