基于AdaBoost的人行横道红绿灯自动识别算法

ALGORITHM OF AUTOMATICALLY RECOGNISING TRAFFIC LIGHT SIGNALS ON CROSSWALK BASED ON ADABOOST

鉴于目前绝大部分人行横道红绿灯没有配合声音提示,利用带摄像头的可穿戴计算设备,通过人行横道红绿灯的自动识别技术,可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算法对红绿灯进行检测定位,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。

Whereas majority of traffic lights over the crosswalks aren＇t equipped with sound alarm system, automatic recognition of traffic light signals on crosswalk will greatly improve the safety of the blind persons crossing the road with the help of wearable computing device with video camera. Integrating cascading AdaBoost and colour filtration, this thesis puts forward an automatic identification algorithm for the traffic lights on pedestrian crossing. First, the method uses AdaBoost algorithm to detect the location of traffic lights. Then, the red lights and green lights will be filtered and classified based on the colour analysis in hue subspace of HSI colour model. The experiments carried out on 418 images of actual complex scenes show that this algorithm is able to achieve quite good detection effect.