摘要
城市道路交通标志检测是现代智能交通的重要组成部分。本文通过改进基于颜色的图像快速分割算法,对交通标志检测信息进行分割压缩处理,以确保交通标志信息分割的完整性和时效性;并依据交通标志在多帧图像序列中的连续变化关系构建空间时序关系模型(Temporal Spatial Model, STM),实现交通标志检测候选区域的筛选处理,以降低交通标志检测误检率、提高准确率。实验结果表明,本文提出的算法有效改进了基于单帧图像的交通标志检测方法存在的误检率高问题,满足复杂的城市道路环境下交通标志检测准确率、实效性和鲁棒性的要求。
Detection of Urban road traffic sign is an important part of the Modern Intelligent Transportation System (ITS). In order to extract the integral areas of traffic signs and reduce the computation load, this paper has improved the fast color-segment compression algorithm. And, it verifies the traffic sign candidate areas to decrease false positives and raise the accuracy by analyzing the variation in preceding video-images sequence while implementing the proposed Spatial Temporal Model (STM). Experimental results indicate that the accuracy, efficiency and the robustness of the framework are better than the method based on one frame, and it can be applied to detect urban road traffic signs in real time.
出处
《计算机科学与应用》
2017年第5期463-472,共10页
Computer Science and Application
