结合颜色分量特征和改进YOLO算法的中国交通标识牌检测

Chinese Traffic Sign Detection Based on Color Component Characteristics and Improved YOLO Algorithm

为了提高利用YOLO算法在自然环境下中国交通标识牌检测的准确性,提出了结合颜色分量特征和改进YOLO深度学习算法。该算法根据中国交通标志图像的颜色特征,获得中国交通标志牌的RGBN空间下的颜色分量特征图;之后,通过在YOLO算法的每个残渣网络中嵌入卷积层的注意力模块(CBAM)改进YOLO算法,进而提升YOLO算法的准确度;最后,利用改进的YOLO算法训练正负样本的RGBN颜色特征图,实现对自然环境下的中国交通标志牌的快速准确检测。实验结果表明:由于论文提出的方法可以去除了图像中大量干扰因素,使得算法在自然环境下检测精度上得到了提升。

To improve the use of YOLO algorithm in the natural environment of Chinese traffic sign detection accuracy,this paper proposes a Chinese traffic sign detection method based on color component characteristics and improving YOLO deep learning algorithm. Firstly,this algorithm takes the color features of Chinese traffic sign images to get color component feature map in RGBN space. Then,embedding the convolutional block attention module(CBAM)into each residue network,the YOLO algorithm can be improved,which can improve the accuracy of YOLO algorithm. Finally,the improved YOLO algorithm is used to train the RGBN color feature map of positive and negative samples to realize the rapid and accurate detection of Chinese traffic signs in natural environment. The experimental results show that the proposed method can remove a large number of interference factors in the image,which improves the detection accuracy of the algorithm.