基于AlexNet卷积神经网络的大米产地高光谱快速判别

采集我国东北和非东北10个产地、4个品种共计1000份单粒大米样本在波长950～1700 nm区间的高光谱图像,按照单粒大米轮廓提取感兴趣区域并计算平均光谱,采用主成分分析从样本集光谱矩阵提取累计贡献率大于99％的第一、二主成分,根据载荷矩阵系数最大值筛选与第一、二主成分相关性最强的特征波长1396.67 nm和146...

Fast Hyperspectral Discrimination of Rice Origin Based on AlexNet Convolutional Neural Network

There were 1 000 single-grain rice samples collected from 10 origins and 4 varieties in Northeast and non-Northeast China.Near-infrared hyperspectral images were acquired within the wavelength range of 950 nm to 1 700 nm.The region of interest was selected according to the outline of single rice from the images to calculate average spectra.Firstly,principal component analysis was used to extract the first and second principal components with a cumulative contribution rate greater than 99%.According to the maximum value of the weight coefficient in the loading matrix,the characteristic wavelengths of the first and second principal components of 1 396.67 nm and 1 467.38 nm are screened respectively.The principal component analysis was performed on the two sets of characteristic wavelength images,and the first three-dimensional principal components were selected respectively,and a total of 2×3 sets of training sample sets can be obtained.Finally,there were 6 models established to discriminate Northeast/ Non-Northeast rice.Among them,the best performing model was built based on the 1 467.38 nm third principal component image,and its recognition accuracy can reach 99.5%.