摘要
根据脐橙图像的特点和分级标准,运用计算机视觉和神经网络算法对脐橙进行自动检测与分级。采用中值滤波和线性低通滤波技术对原始脐橙图像进行平滑、去噪,在对脐橙图像像素点颜色信息统计的基础上,通过设置蓝色分量、色调、饱和度的阈值,从图像中快速准确的分割出果实图像;确定果实横径、果形、表面缺陷率、色泽与着色率为脐橙外部品质分级的特征参数;通过BP神经网络建立了特征参数与脐橙等级之间的关系模型,试验结果表明,其预测准确率达到85%。
According to image characteristics and classification standards of navel orange, computer vision and pattern recognition technology was used to realize automatic detection and classification. The original image was disposed fast and smoothly by median filtering and linear low passing filtering. Based on the statistical treatment of picture element color information of the navel orange image, the fruit image was effectively wiped off by setting up the values of B （blue）, H（hue） and S（saturation）. The parameters of diameter, shape, surface defects, color, and pigmentation ratio were extracted according to grading standards. The model referring the relation between character parameters and navel orange grading was set up by BP neural network. The results showed that the forecasting nicety could reach 85%.
出处
《中国农业科技导报》
CAS
CSCD
2008年第4期100-104,共5页
Journal of Agricultural Science and Technology
基金
国家自然科学基金项目(30460059)资助
关键词
计算机视觉
脐橙
特征参数
BP神经网络
computer vision
navel orange
character parameter
BP neural network
