Abstract: Hyperspectral is widely used in agriculture, and   it provides certain technical support for the classification and identification of crop varieties due to fast, efficient, accurate and non-destructive characteristics. This paper collected canopy reflectance spectra data of 11 rapeseed varieties in the pre-seedling and late seedling stages, using a total of 23 characteristic parameters in 6 aspects of hyperspectral position, amplitude, area, width, reflectance and vegetation index as indicators to measure the contribution rate of the characteristic parameters and the significance of the analysis of variance. Based on this, the effect of distinguishing and identifying on different varieties of rape was analyzed. The results showed that: from the view of contribution rate, different types of hyperspectral characteristic parameters had different ability to distinguish rape varieties, the amplitude parameter had the strongest ability to distinguish rape varieties, and the width parameter had the weakest ability to distinguish rape varieties. The overall effect from strong to weak was: amplitude parameter>area parameter>reflectance parameter>vegetation index parameter>position parameter>width parameter; from the perspective of analysis of variance, the effects of distinguishing rape varieties in different periods were different. Among them, the comprehensive effect of identifying rape varieties with vegetation index parameters was the best. The identification effect in the late seedling stage was better than pre-seedling stage, the three characteristic parameters of D_r, SD_r/SD_{b and} (SD_r-SD_y)/(SD_r+SD_y) were the best in the multiple comparison analysis of variance, which could clearly distinguish 6 varieties. Canopy spectral characteristic parameters at seedling stage could classify and identify rape varieties. The research results laid the foundation for the rapid classification and identification of crops and the rational formulation of the planting area and spatial distribution of crops.

Key words: rape, variety identification, characteristic parameters, seedling stage, variance analysis

摘要： 高光谱技术在农业领域应用广泛，其快速高效、精准无损的特点为农作物品种分类识别提供了一定的技术支持。采集11个油菜品种苗前期、苗后期冠层反射光谱数据，以高光谱位置、振幅、面积、宽度、反射率和植被指数6个方面共23个特征参数为研究指标，衡量特征参数的贡献率大小和方差分析显著性，据此分析其区分、识别油菜不同品种的效果。结果表明：从贡献率的角度，不同类别的高光谱特征参数区分油菜品种的能力不同，振幅参数区分油菜品种的能力最强，宽度参数区分油菜品种的能力最弱，综合效果由强到弱依次为：振幅参数＞面积参数＞反射率参数＞植被指数参数＞位置参数＞宽度参数；从方差分析的角度，不同时期区分油菜品种的效果不同，苗后期的识别效果优于苗前期，其中，以植被指数参数识别油菜品种的综合性效果最好。多重比较方差分析中以D_r、SD_r/SD_b、(SD_r-SD_y)/(SD_r+SD_y)3个特征参数识别效果最优，可以明确区分6个品种。苗期利用冠层光谱特征参数能够较好的对油菜品种进行分类识别，研究结果为快速实现农作物的分类识别及合理制定农作物的种植面积和空间分布奠定了基础。

关键词: 油菜, 品种识别, 特征参数, 苗期, 方差分析