基于Hamming网络的苹果颜色分级研究  

Study on Color Sorting for Apples Based on Hamming Neural Networks

作　　者：胡海晴[1] 邓继忠[1] 张泰岭[1] 

机构地区：[1]华南农业大学工程学院,广州510642

出　　处：《农机化研究》2006年第11期75-76,80,共3页Journal of Agricultural Mechanization Research

基　　金：广东教育厅自然科学研究项目(粤教科[2001]72号)

摘　　要：在基于计算机视觉的水果分选中,颜色是最主要的指标之一。为此,将Hamming神经网络应用于苹果的颜色分级。依据富士苹果的颜色将其分为3个等级,分析了各等级的平均色调累积直方图,据此构建了Hamming神经网络;从直方图上提取色调在0o,10o,20o,30o,40o,50o所对应的频度作为Hamming网络的输入向量;采集了3个等级各10个苹果图像作为训练样本,另外同样30个苹果图像作为测试样本。试验结果表明,Hamming网络颜色分级的准确率不低于96%,识别1个苹果的时间在0.1～0.2ms之间。Color is one of the primary index in the sorting fruit based on comport vision. In this paper the Hamming neural network was applied to apples color sorting. First, Fuji apples were sorted into three classes according to their color conditions, and the average accumulating histogram of the hue value about the three classes were analyzed. Based on the analysis, then the framework of Hamming neural network was established. The accumulating Frequency of the hue value at 0° , 10° , 20° , 30° , 40° , 50° degrees were used as input vector of the neural network classifier. 30 apples （ 10 apples for each class） were used for neural network training, another 30 apples were used for testing. The result shown that the classification accuracy is over 96% and the speed is between 0.1 and 0.2ms for each apple

关 键 词：计算机应用 颜色分级 理论研究 计算机视觉 苹果 HAMMING神经网络 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]