Abstract <p>Color sorters demonstrate high efficiency in cleaning seed material from contaminants. However, when processing small-seeded crops, the resolution of color sorter cameras does not ensure the effective detection of hard-to-separate impurities. The study aimed to determine the optimal settings for the SmartSort C color sorter systems for removing quarantine dodder seeds during the final cleaning of the Artemida alfalfa variety harvested in 2024. The authors used the following operating parameters of a color sorter to sort samples and subsequently analyze the cleaned purified fraction and waste: an ejector response delay of 2.6, 27, and 3.3 ms, a pressure of 1 and 2 bar, and illumination of white light and white light combined with infrared (IR) radiation. The efficiency of dodder seed separation was determined for alfalfa seed contamination of 0.63, 0.32, and 0.14%. When comparing the images from the RGB cameras of the color sorter obtained with IR illumination together with white light against the background of the IR spectrum, it was found that alfalfa seeds with a smoother shell texture are overexposed and look like white spots, while dodder seeds with a rough shell contrast with black dots. When IR illumination was combined with white light, the contrast of dodder seeds increased by 40%. An increase in dodder seed contamination by 0.1% leads to a decrease in the quality of alfalfa seed cleaning by the color sorter, on average, by 1.16%. The complete removal of dodder seeds from alfalfa seeds by means of an IR background and white light illumination was observed at a pneumatic system pressure of 2.0 bar, an ejector response delay of 2.6 milliseconds, a specific machine load per tray of 85 kg/h, and a contamination level of 0.14%. The study results can be used in the implementation of machine learning and artificial intelligence systems in the software of domestic color sorters.</p>