Improved traffic signs detection based on significant color extraction and geometric features

Traffic signs detection is a key part of traffic signs recognition system. We propose an improved approach for traffic signs detection based on significant color extraction and geometric features. Firstly, we use a median filter to remove the noise. Secondly, we calculate the quadratic weighting difference of R, G and B in RGB color space to extract the significant color of the traffic sign. Thirdly, the morphological processing is applied to get connected regions. Finally, we filtrate the connected regions based on geometric features to locate the traffic sign accurately. The experiment on 200 traffic sign images shows the detection rate of 95.4%. Our algorithm can adapt to various environment, has a high accuracy and a good robustness.