甘蓝型油菜粒色近红外光谱分析模型构建  

Construction of near-infrared reflectance spectroscopy model for seed color of rapeseed

作　　者：李玉[1] 刘勋[1] 李加纳[1] 殷家明[1] 徐新福[2] 

机构地区：[1]南方山地农业教育部工程研究中心,重庆北碚400716 [2]西南大学,重庆市油菜工程技术研究中心,重庆北碚400716

出　　处：《中国油料作物学报》2012年第5期533-536,共4页Chinese Journal of Oil Crop Sciences

基　　金：高等学校学科创新引智计划(B12006);重庆市攻关项目(2010GGC018)

摘　　要：以306份甘蓝型油菜种子为研究材料,通过计算机图像处理技术筛选出最能反映油菜籽粒色变化的RGB色彩空间参数,并随机选择246份材料对粒色信息中的R(红色)值建立近红外光谱分析模型。在油菜粒色数字信息中,R值最能体现油菜籽粒色性状的变化情况,不同粒色等级对应样品的R值为:0级小于55、1级在55～100之间、2级在100～125之间、3级在125～140之间、4级在140～150之间、5级大于150。对R值构建NIRS分析模型,12种数学及光谱处理组合中最优模型的内部交叉检验相关系数为0.900;外部检验相关系数为0.831,检验偏差为2.147。说明利用RGB色彩空间下的R值构建粒色近红外光谱分析模型是可行的,该模型具有较高的稳定性和准确性,可以作为油菜种子粒色的定性和定量分析的依据。Accessions of rapeseed（Brassica napus L.） were assessed to select an RGB color definition model representing the seeds color variation by computer image processing.Out of total 306 accessions,246 were randomly selected to construct the near-infrared reflectance spectroscopy（NIRS） analysing model according to R（red） value.Seed color was evaluated by different grades of R values.In Grade 0,R values were below 55,others were Grade 1 with R between 55 and 100,Grade 2 with R valure between 100 and 125,Grade 3 with R value between 125 and 140,Grade 4 with R value between 140 and 150,and Grade 5 with R value with R value greater that 150.The NIRS model for R value was constructed by optimizing 12 combinations of mathematical and spectral treatments.The optimal model had internal validation of 0.900 and external validation 0.831 respectively,and the test deviation was 0.831.The result showed that using R value of RGB color space was feasible to build NIRS model with high accuracy and stability and could be used for qualitative and quantitative analysis for rape seed color.

关 键 词：油菜 粒色 近红外光谱 分析模型 

分 类 号：TN219[电子电信—物理电子学] S565.4[农业科学—作物学]