A PCA-Based Material Color Classification Method via LED Sensing

Considering the low implementation cost, this work seeks to classify the material color by adopting one polychromatic sensing light-emitting diode (LED) array. The color classification problem is equivalent to determining the spectral reflectance vector based on the output of the sensing LED array. For robust color classification, the generalized likelihood ratio test method is employed to jointly estimate the observation scalar and detect the spectral reflectance vector. Based on the principal component analysis, the detection complexity and the required number of sensing LEDs can be greatly reduced by transforming the detection of the high-dimensional spectral reflectance vector into the detection of a low-dimensional weight vector. The experimental results demonstrate that the proposed method can achieve a high classification accuracy of nearly 100% and is robust to the link gain variation.