【摘要】 以入侵植物薇甘菊高光谱图像为研究对象,基于4种预处理方法对薇甘菊高光谱图像进行降低噪声处理,分别研究了基于主成分分析的特征提取方法和基于BP神经网络的分类模型,筛选出薇甘菊高光谱识别的最优预处理方法,以实现薇甘菊的快速准确识别。结果显示,预处理方法为一阶、二阶微分的识别率分别为81.2%和76.92%;标准正态变量变换(SNV)和一阶微分+SG平滑的识别率分别为89.74%和87.18%。多次试验得到基于SNV预处理方法的识别率最稳定,即得到最优预处理方法为SNV。更多还原

【Abstract】 With invasive plants Mikania mikania Kunt hyperspectral image as the research object,based on four kinds of pretreatment method of M.micrantha hyperspectral image noise reduction processing,we studied the feature extraction method based on principal component analysis and the classification model based on BP neural network.The optimal pretreatment method for M.micrantha hyperspectral identification was screened,in order to realize fast and exact recognition of M.micrantha.The experimental results showed that the recognition rates of first and second orders were 81.2% and 76.92%,respectively.The recognition rates of standard normal variable transformation(SNV) and first order differential +SG smoothing were 89.74% and 87.18%,respectively.Multiple experiments showed that the recognition rate of SNV-based pretreatment method was stable,in other words,the optimal pretreatment method was SNV.更多还原