By researching current road signs recognition methods, we found that SIFT is rarely used for real time system because its computation cost is high; while PCA-SIFT often fails when objective images have obvious changes in scaling and rotation. To solve the problems of SIFT and PCA-SIFT, we propose a more effective real time road signs recognition algorithm named HSI-SIFT, it combines HSI image segmentation with SIFT recognition. We illustrate that HSI-SIFT can be used for real time road signs recognition efficiently; and it is more accurate and robust than PCA-SIFT to changes in scaling, rotation and illumination; moreover, HSI-SIFT is about 10 times faster than PCA- SIFT on average.