基于高光谱图像技术的菜用大豆厚度检测

Thickness Measurement of Green Soybean Using Hyperspectral Imaging Technology

DOI：10.3969/j.issn.1673-1689.2012.11.004

中文关键词: 菜用大豆 高光谱图像 厚度检测 建模

英文关键词: vegetable soybean hyperspectral images thickness measurement modeling

基金项目: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)

摘要点击次数: 204

全文下载次数: 736

中文摘要:

菜用大豆厚度是划分菜用大豆等级的重要衡量指标之一。采用高光谱图像技术对菜用大豆的厚度进行预测。实验中选取200个菜用大豆作为测试样本,获取其高光谱反射图像,同时用数字式游标卡尺测量厚度值。选取400～1 000 nm范围的光谱信息,采用多元散射校正、标准归一化和导数计算对光谱数据预处理,结合偏最小二乘和多元线性回归两种分析方法建立厚度校正模型和预测模型。研究发现基于多元散射校正的偏最小二乘方法的模型精度较优,校正模型和预测模型的相关系数分别为0.956和0.933,均方根误差分别为0.59 mm和0.70 mm。研究结果表明可以利用高光谱图像技术预测菜用大豆厚度。

英文摘要:

Thickness is an important index for the grade of green soybean.In this manuscript,hyperspectral imaging technology was used to determine the thickness of two hundreds green soybean.The hyperspectral reflectance images of the samples were acquired using hyperspectral imaging system and the instrumental thickness values were measured by digital caliper.The range of 400 ～1000 nm spectral information was preprocessed using multiplicative scatter correction algorithm(MSC),standard normal variate algorithm(SNV) and derivation calculation algorithm,respectively.Then,partial least squares(PLS) and multiple linear regressions(MLR) were used to develop calibration and prediction model.Results show that optimal models are obtained by MSC coupled with PLS method,which the correlation coefficients of calibration set and prediction set were 0.956 and 0.933 and the root mean square error is 0.59 mm and 0.70 mm,respectively.The aboved results demonstrated that hyperspectral imaging technology is suitable for detection of green soybean thickness e.

查看全文 查看/发表评论 下载PDF阅读器