In the research of segmentation of the target image for tomato harvesting robot, the accuracy was affected by illumination and calculation complication. Firstly, the characteristics of RGB, HIS and YUV color spaces were compared. It was theoretically analyzed that the YUV color space applying to tomato harvesting robot vision system was feasible. Then, the threshold value of the ripe tomato in their color spaces was given according to the results of experiments and experience. The double-threshold algorithm based on YUV color space was put forward to apply to the tomato harvesting robot vision system. And the experiments under different illumination were designed to segment ripe tomato respectively in RGB, HIS and YUV color space in order to test the effect of those algorithms. The experiments also showed that the algorithm based on YUV color space had the better robust and the better real-time than the others. 