基于改进SSD的道路交通标志检测

Road Traffic Sign Detection Based on Improved SSD

针对复杂环境下交通标志检测精度低的问题,设计了一种检测精度更高的目标检测算法,对SSD深度学习目标检测算法进行了优化改进;将深度特征表征能力较强的Resnet50网络模型融入于SSD算法中;采用K-means++聚类算法确定SSD中先验框的尺寸,提高交通标志的检测率;分别利用SSD模型和改进的SSD模型做检测对比实验,结果表明,改进算法对各类型交通标志的检测精度比原SSD算法更高;改进的SSD方法对交通标志进行检测能取得较好效果,弥补了原算法的不足。

Aiming at the problem of low detection accuracy of traffic signs in complex environments,a target detection algorithm with higher detection accuracy is designed,and the SSD deep learning target detection algorithm is optimized and improved.Incorporate the Resnet50 network model,which has strong deep feature characterization capabilities,into the SSD algorithm.The K-means++clustering algorithm is used to determine the size of a priori box in the SSD to improve the detection rate of traffic signs.The SSD model and the improved SSD model were used to do detection comparison experiments.The results show that the improved algorithm has higher detection accuracy for various types of traffic signs than the original SSD algorithm.The improved SSD method can achieve better results in the detection of traffic signs,which makes up for the deficiencies of the original algorithm.