基于改进SSD的道路交通标志检测

Road traffic sign detection based on improved SSD

摘要:

交通标志检测在无人驾驶领域有重要的应用。针对复杂环境下交通标志检测精度低的问题，本文对SSD算法进行优化改进，提出将特征提取网络从VGG16替换为特征提取能力更强的Resnet50，为了提高检测效果，本文采用K-means++聚类算法确定SSD中先验框的尺寸。本文基于TensorFlow深度学习框架，对复杂环境下的图像中的交通标志进行定位和分类，分别利用SSD模型和改进的SSD模型做检测对比实验，分析模型测试结果。结果表明，改进方法对各类型交通标志的检测精度相比原SSD算法更高。通过本文的方法对交通标志进行检测能取得较好效果。

Abstract:

Traffic sign detection has important applications in the field of unmanned driving. Aiming at the problem of low detection accuracy of traffic signs in complex environments, this paper optimizes and improves the SSD algorithm, and proposes to replace the feature extraction network from VGG16 with Resnet50 with stronger feature extraction capabilities. In order to improve the detection effect, this paper uses the K-means++ clustering algorithm Determine the size of the a priori box in the SSD. Based on the TensorFlow deep learning framework, this paper locates and classifies traffic signs in images in complex environments, and uses the SSD model and the improved SSD model to perform detection and comparison experiments, and analyze the model test results. The results show that the improved method has higher detection accuracy for various types of traffic signs than the original SSD algorithm. The method in this paper can achieve better results in detecting traffic signs.