The illegal action of exceeding the powdered food additives standard seriously threatens the health of people. In the study, a method for qualitative and quantitative prediction of azoformamide doping in flour through Raman hyperspectral images was developed by using a self-built Raman hyperspectral detection system. In this method, Raman hyperspectral images near 785 nm of azoformamide in samples were obtained through laser line light source. By preprocessing, selecting the region of interest, data dimensionality reduction and setting an appropriate threshold, the effective distinction between flour and azoformamide signals is realized. The method of image analysis was used to detect the azoformamide doped in samples with gradient concentration. Then the related quantitative analysis model was established. Finally the reliability of the quantitative analysis model is verified by the prediction sets and the correlation coefficient is more than 0.988. This study provides a new method for the detection of powdered food by Raman hyperspectral technology.