摘要
形状判别是苹果外观品质检测中不可缺少的内容。本文先后采用主动形状模型(ASM)和基于傅立叶描述子的神经网络方法进行苹果形态分级。实验结果表明:传统神经网络方法的判别准确率为83．3%左右,而ASM方法的分级效果较好,对苹果果形的判别准确率高达95%,模型与实际对象匹配的时间不超过2s,且直观性强、鲁棒性好,具有较好的灵活性,能够满足苹果实时分级的需要。
Apple shape identification is an essential character on appraising its appearance quality. This paper introduces a method of active shape models （ASM） as the neural network method based on Fourier to identify the apple shape. The experiment results demonstrate that the accuracy of the neural network is about 83.3% and the ASM method has good effect reaching as high as 95%. The matching time between the model and the actual image does not surpass 2 see. ASM has good visibility, high flexibility and strong robustness to satisfy the real-time graduation of apple.
出处
《食品科学》
EI
CAS
CSCD
北大核心
2007年第4期56-59,共4页
Food Science
基金
江苏省高校自然科学基金项目(05KJB210019)
关键词
苹果
果形
ASM
主成分分析
BP
傅立叶变换
apple
shape
ASM
principal component analysis
BP（error back propagation）
Fourier transformation
