The paper aims to realize efficient characterization of the LCD display RGB color space to color spectrum. The principal component analysis method was used to reduce the dimensionality of the spectral data and the RBF neural network was used to study the influence of the input variable data range, visual weighting function and color quantity on the accuracy of the characterization model. The original information of the spectrum could be properly preserved when the number of principal components was 6. The characterization accuracy was high when the number of principal components was 6, the input variable range was 0 to 2.55, the CIE1931 visual function was used as the weighting function, and the number of colors was 364. The average color difference of 99 color conversions was objectively verified to be 0.36, the maximum color difference was 1.59, and the average color difference of all color patches was 0.17. The input variable data range has the greatest impact on the model, and the weighting function and the number of colors are the second. Therefore, the input variable range, visual weighting function and number of colors should be considered in the characterization to improve the accuracy of the model. The model proposed is a high-precision characterization model with certain practical application value.