基于感兴趣区域提取与双过滤器的交通标志检测算法

Traffic sign detection algorithm based on region of interest extraction and double filter

为了解决当前交通标志检测(TSD)技术因各种复杂因素的干扰,导致其难以有效对交通标志进行正确检测,以及鲁棒性较弱等不足,设计了一种基于感兴趣区域提取与双过滤器的交通标志检测算法。首先,为了降低环境干扰,对输入图像进行预处理,以增强每个标志的主颜色。其次,为提高对感兴趣区域提取能力,定义了基于最大稳定极值区域(MSER)与波动方程(WE)的感兴趣区域(ROI)检测器,通过ROI检测器选取候选区域。然后,引入有效的方向梯度直方图(HOG)描述符作为交通标志检测特征,并利用支持向量机(SVM)进行分类,将其划分为交通标志或背景。最后,利用上下文感知过滤器与交通灯过滤器进一步识别伪交通标志,提高检测精度。在German交通标志数据库(GTSDB)中对常见的指示性、禁止性和危险性的3种交通标志进行测试,结果表明,与当前交通标志识别技术相比,所提算法对交通标志检测具有更高的检测正确率与鲁棒性。

In view of the complex environmental impact on traffic sign detection,it was difficult to detect the traffic signs correctly,and the robustness was weak,a traffic sign detection scheme based on the extraction of region of interest( ROI) and coupling dual filter was designed. Firstly,to reduce the environmental interference,the input image was pretreated to further enhance the main color of each logo. Secondly,to improve the extraction ability of regions of interest,a complementary maximally stable extremal regions( MSER) and wave equation( WE) ROI detector were defined,the candidate regions were selected by ROI detector. Then,the introduction of effective histogram of oriented gradient( HOG) as a traffic sign detection feature descriptor,and the use of support vector machine( SVM) are classified into traffic signs or background. Finally,through the context aware filter and traffic light filter to further clarify the pseudo traffic signs,the detection accuracy is improved. Through the test of three kinds of traffic signs which are indicative,prohibitive,and dangerous in GTSDB database,the results show that the proposed algorithm has excellent accuracy and robustness for traffic sign detection.