柑橘花钾素营养的高光谱表征

Characteristics of potassium content in citrus flowers with hyperspectral imagery

【目的】试验以枳橙砧‘卡拉卡拉红肉脐橙’花器为试材,利用高光谱技术对柑橘花钾素营养进行表征,以期为柑橘植株营养的信息化管理提供理论依据。【方法】探索通过单波长阈值分割法提取柑橘花高光谱图像信息,利用PLS建模筛选对柑橘花钾素含量响应最优光谱数据,以此数据建立i PLS和si PLS的钾素含量估测模型,筛选出特征波段及组合,建立柑橘花钾营养的高光谱估测技术。【结果】采用花朵高光谱扫描信息,设置670.66 nm光谱0.5~0.6灰度区间可以较好提取到柑橘花组织有效光谱信息,筛选出558.17~612.16 nm特征波长和558.17~612.16 nm、612.16~666.69nm、666.69~721.69 nm特征波段区间组合,柑橘花钾素含量估测si PLS模型的预测效果最佳,其校正集相关系数和均方根误差分别为0.954 7和0.089 0,预测集相关系数和均方根误差也分别达0.885 5和0.161 9。【结论】建立的柑橘花钾素含量估测有效信息提取方法、特征光谱以及最优预测模型,为柑橘花钾素营养实时检测奠定了基础。

【Objective】The hyperspectral imaging technique was developed to realize a fast,accurate andobjective detection of potassium content in citrus flowers,providing a theoretical basis for information man-agement of citrus.【Method】Firstly,the hyperspectral images of citrus flowers were collected. Then,a suit-able method was developed to extract the effective region from the hyperspectral image of the flower,whichset the feature wavelength(670.66 nm) for potassium content in flowers,and the grey image were segmentedby different gray thresholds with software. The calculated average spectrum of each gray threshold rangewas used to established the calibration models of citrus flower potassium content employing partial leastsquares(PLS). The gray range(0.5~0.6) was selected by comparing these parameters of the models(Rc=0.908 3 and RMSECV= 0.124 9). Then Interval partial least square(i PLS) and synergy interval partial leastsquare(si PLS) methods were respectively employed to choose the characteristic wavelength and spectralbands combination based on the selected gray range(0.5~0.6)data.【Result】The results showed that si PLSmodel was obtained with 11 intervals,the optimal combinations of intervals selected were No.4,No.5,No.6,with Rc=0.954 7 and RMSE CV=0.089 0 in calibration set,and Rp=0.885 5 and RMSEP = 0.161 9 in vali-dation set.【Conclusion】The results illustrated that citrus flower potassium content could be nondestruc-tively and real-time estimated by the low cost technique named hyperspectral imaging.