中文摘要：

目前基于机器视觉的检测方法应用于马铃薯形状识别相对较少,且多集中在矩特征等方法中,计算量大。为此,利用光学相关联合变换的理论,实现了马铃薯形状的检测和识别。该方法利用相关联合变换中所形成的旁瓣峰值来表明物体相似度情况并作为评价薯形相似性的依据,利用该特征实现马铃薯形状的相关性检测和形状识别。实验表明,随着图片库样本的增加,该方法提高了薯形的识别率。当样本库图片为300幅时,对于圆形、椭圆和畸形的识别率分别为91.25%,93.75%,88.75%,且对样本马铃薯的识别率为100%。该方法对于图片的要求很低,可广泛使用于马铃薯收获和食品加工环节,且分类准确率高,具有一定的实用价值。

英文摘要：

At present,test method based on machine vision has been applied to the quality test of agricultural products widely.But the related study on potato shape identification was very little.In this article,theory of the optical joint transform correlation was quoted,and MATLAB＇s scientific computing and drawing functions were used to recognize potatoes＇ shapes accurately.The recognition rate of potato shape has increased when the samples picture increased.When the image database increased to 300,the recognition rate of the round,oval and malformation potato result in 91.25%,93.75% and 88.75% respectively.Its detection accuracy can reach 100% for the samples pictures.Compared with optical system test,this method has many advantages,such as low requirement to the pictures,high recognition rate and possesses practical value.Therefore this method can be widly used in potato harvest and potato food processing.