摘要
提出了一种利用高光谱图像技术对玉米种子产地和年份的鉴别方法.首先采用高光谱成像系统采集不同产地和年份的玉米种子高光谱图像,利用主动轮廓模型对玉米种子高光谱图像进行轮廓提取,得到每粒玉米在400~1 000 nm共233个波段范围内的4个光谱特征,利用不同的特征及预处理方式结合偏最小二乘判别分析建立玉米种子的产地和年份鉴别模型.结果显示,利用最佳特征及预处理方式建立的玉米种子产地和年份鉴别模型中,训练集和测试集精度分别为99.11%和98.39%.研究结果表明,利用高光谱图像技术对玉米种子的产地和年份进行无损鉴别是可行的.
Hyperspectral image technology was investigated to identify the geographical origin and years of maize seeds.First,hyperspectral images of different geographical origin and years were acquired using hyperspectral imaging system.Subsequently,four spectral features of each maize seed covered the spectral region of 400 ~1 000 nm contained 233 wavelengths were extracted utilizing the results of active contour model.Final,the identification models were developed using different spectral characteristics and preprocessing methods coupled with partial least squares discriminant analysis.The result showed that the accuracy rate of training and testing set was 99.11%and 98.39% for geographical origin and years identification using optimal feature and preprocessing method,respectively.It is indicated that hyperspectral image technology is an effective method to identify the geographical origin and years of maize seeds.
出处
《食品与生物技术学报》
CAS
CSCD
北大核心
2014年第2期163-170,共8页
Journal of Food Science and Biotechnology
基金
国家自然科学基金项目(61271384
61275155)
江苏省自然科学基金项目(BK2011148)
中国博士后基金项目(2011M500851)
关键词
高光谱图像
玉米种子
产地
年份
偏最小二乘判别分析
hyperspectral image
maize seeds
geographical origin
years
partial least squares discriminant analysis