The invention discloses a qualitative detection method for an illegal ingredient, i.e., Sudan red, in raw materials of feeds. According to the method, suspected areas are primarily screened through microscopic image processing, so a detection scope is reduced; then a microscopic Raman imaging spectrometer is used for detecting the Raman spectra of the suspected areas, the spectra are pretreated so as to obtain training samples, an IVISSA method sensitive to classes is used for waveband selection, and the IVISSA method can select characteristic peaks of Sudan red substantially different from characteristic peaks of other components, which enables detection efficiency and accuracy to be improved; the average spectrum of each training sample is taken and incorporated into a standard spectrum matching database; and the microscopic Raman spectrum of an unknown sample is acquired according to coordinates of detection points obtained in primary image election, then same spectral pretreatment is carried out, and a same wavelength variable is chose for one-to-one matching with each spectrum in the standard spectrum matching database according to a feature enhanced spectral angle matching method so as to determine whether the unknown sample contains Sudan red or not. The qualitative detection method has the advantages of rapid detection, good stability, and lower false positive rate and detection limits.