Near infrared spectral technology (NIR) was used to test talc-containing wheat flour. The spectrum was preprocessed with multiplicative scatter correction. The quantitative analysis model of talc-containing flour was built using SCG back propagation algorithm training function of BP neural network, and the calibration set and prediction set were quantitatively analyzed. R2 was 0.9973, the root mean square error of calibration (RMSEC) was 0.4367, and the root mean square error of prediction (RMSEP)was 1.7088. The results showed that BP neural network with NIR for the determination of talc-containing flour has the advantages of fast, high precision, and the ability of Fanhua, and can be used for talc-containing flour.