Hyperspectral Image Classification for Foreign Bodies in Medicine Based on Spatial-Spectral Fusion Feature

As the cornerstone of the pharmaceutical industry and the important guarantee for people’s health, pharmaceutical safety has attracted more and more attention from both inside and outside the industry, in which the detection of foreign bodies has always been an important link. In view of the advantages of hyperspectral technology in material classification, a new hyperspectral classification method for foreign body detection in medicine based on spatial-spectral fusion feature is proposed in this paper, aiming at the problems of "dimension disaster" and low fusion degree of spectral feature and spatial feature. Firstly, a method combining PCA dimension reduction of band clustering grouping with semi-supervised LDA is proposed to extract spectral feature, while 2D Gabor filter is used to extract spatial feature. Then, a feature fusion network based on attention mechanism is proposed to fuse the spectral feature and spatial feature. Finally, the full connection layer and softmax classifier are used to realize the detection and classification of foreign bodies in medicine. The proposed method is applied to our own medicine data set, and the experiments constantly prove the effectiveness and superiority of our method.