Aiming at the problem of low grading accuracy caused by illumination non-uniformity of light source in an intelligent grading equipment for flue-cured tobacco, an optimization solution to improve the non-uniformity was proposed. Firstly, a 54-point measurement method was proposed to measure the illumination non-uniformity of the light source, and the problem of illumination non-uniformity of the light source of the used equipment was analyzed. Secondly, the optimization solution of LED line light source and LED surface light source was proposed, and the variance, standard deviation and coefficient of variation of light source illumination were calculated, and the illumination uniformity was analyzed. Finally, the LightGBM model was used to verify the rationality of the optimization solution. The results showed that the illumination uniformity and image sample grading accuracy of the optimized light source were better than those of the old light source to meet the accuracy requirements of the actual acquisition of flue-cured tobacco.