Abstract: Whereas the majority of traffic lights for crosswalks aren’t equipped with sound alarm system，automatic recognition of traffic light signals on crosswalk will greatly improve the safety of the blind crossing the road，using moble phone with a camera.Integrating cascade Adaboost and color filtration，this paper puts forward an automatic identification calculating method for the traffic lights on pedestrian crossing.Firstly，the method uses Adaboost calculation to detect the location of traffic lights.Then，the red lights and green lights will be filtered and classified based on the color analysis in the hue subspace of HSI color model.The experiments on the 418 images of actual complex scene show that this calculating method has achieved great success.

Key words: traffic lights, traffic sign recognition, image detection

摘要： 鉴于目前绝大部分人行横道红绿灯没有配合声音提示，利用带摄像头的智能手机，通过人行横道红绿灯的自动识别技术将大大提高盲人过马路的安全系数。提出一个融合级联Adaboost与颜色过滤的人行横道红绿灯自动识别算法。该方法首先利用AdaBoost算法对红绿灯位置进行检测；然后在HSI色彩模型的色调子空间进行颜色分析的基础上进行红绿灯的过滤与分类。用采集的418幅实际复杂场景下的图像数据进行实验，结果表明该方法能达到较好的检测效果。

关键词: 红绿灯交通标志, 道路交通标志识别, 图像检测