基于高光谱成像技术的不同类型小麦穗发芽籽粒识别

Identification of Preharvest Sprouting Grains of Different-Type Wheat Varieties Based on Hyperspectral Imaging Technology

【目的】为探索利用高光谱成像技术识别不同类型小麦籽粒穗发芽的方法。【方法】利用涵盖不同倍性、籽粒颜色的3个品种的小麦种子为材料,通过设置穗发芽对比实验,利用PCA、ICA、HSV 3种变换方法分析不同类型小麦正常和穗发芽籽粒光谱、图像的差异,提取图像和光谱特征参数,并运用最小距离法、最大似然法和支持向量机3种不同识别算法进行小麦籽粒发芽识别。【结果】穗发芽与否会在其光谱特性上有所反映,发芽籽粒的光谱反射率在品种间差异较小,发芽与未发芽籽粒的光谱反射率存在较大差异,大于未发生穗发芽品种之间的差异,特别是在光谱波段470~620 nm之间。通过光谱差异分析并结合支持向量机算法进行识别得到的结果精度更高,识别精度达到96%。【结论】研究结果可为种子质量监控、自动筛选识别等提供技术支持。

【Objective】The aim of this study was to explore the method of identifying preharvest sprouting grains of different-type wheat by hyperspectral imaging technology.【Method】Seeds of different ploidy wheatand grain colors were used as materials.The difference between normal and preharvest sprouting grains of different wheat types on spectra and images was analyzed under the condition of grain germination experiment.After extracting the image and spectral characteristic parameter,a total of three differentrecognition algorithm including the minimum distance method,maximum likelihood method and support vector machine were used to identify preharvest sprouting grains.【Result】The spectral characteristics between normal and sprouting grains were different.The spectral reflectance of sprouting grains had little difference between different wheat types,and that difference between normal and sprouting grains on spectral reflectance was much greater than that difference between different wheat varieties without sprouting,especially in the spectral band between 470 nm and 620 nm.Through the spectral difference analysis combined with support vector machine algorithm,we obtained a higher recognition accuracy of 96%.【Conclusion】The results can provide technical support for seed quality control,automatic screening and identification.