Material Intelligent Identification Based on Hyperspectral Imaging and SVM

With the rapid development of economic globalization, the method of material identification in customs is behind the times. The higher identification efficiency is required by the increase of imports. In this paper, a new hyperspectral imaging intelligent material identifying system based on instantly scanning technique is proposed. The system is investigated by an identification experiment with 5 kinds of flour participated. The spectrum imaging device adopted is from MeiyaPico coded CP-2000B. 250 groups of spectral feature (400 to 750 nm) data are collected by the device. Sample sets retreated by removing the envelope using the radial basis function for kernel are transformed into a subspace. The database building and the identification are accompanied by SVM classification algorithm. In the experiments, the accuracy of system proposed is above 95%.