摘要
提出了一种根据苹果气味对苹果进行无损检测的新方法,研制了一套适合苹果气味检测的电子鼻系统。对超市所购得的好坏苹果各50个进行了检测,在获得传感器阵列数据的基础上,从每个传感器曲线中提取了5个特征参数,将其作为模式识别的输入向量。由主成分分析对所测的数据处理结果看出,好坏苹果是可以区分的,但有一点重迭的地方。用遗传算法优化RBF神经网络,发挥各自的优点,使所建立的遗传RBF网络不但收敛速度快,而且识别精度高。网络对训练集的回判正确率和对测试集的测试正确率分别为100%和96.4%。试验证明该分类方法和电子鼻装置都是有效的,也适用于其他的水果。
In this article, a new method to classify apples by the odor of apples is given, and an electronic nose equipment to classify apples is developed. 50 good apples and 50 bad apples bought from the super-market is classed. 5 features parameters are developed from every data curve of sensor arrays, and all the feature parameters are called input vectors. Principal component analysis (PCA) and genetic algorithm radix based function neural network (GA-RBF) are used to combine the optimum feature parameters. Good separation among the gases of different apples is obtained using principal component analysis but a bit is overlapped. The recognition probability of the GA-RBF to the learning samples and the testing samples are 100% and 96.4%. Both this method and the electronic nose equipment are successful from the above experiments. The new method can also be applied to other fruit classification.
出处
《农机化研究》
北大核心
2004年第3期179-182,共4页
Journal of Agricultural Mechanization Research
基金
江苏省自然科学基金项目(BK2001088)
关键词
电子鼻技术
苹果
质量评定
气体传感器阵列
aoto-control technology
electronic nose
experimental
gas sensor array
genetic algorithms
pick up parameters
RBF network
