Journal of Frontiers of Computer Science and Technology ›› 2017, Vol. 11 ›› Issue (1): 134-143.DOI: 10.3778/j.issn.1673-9418.1607035

Previous Articles     Next Articles

Moving Vehicles Recognition in Complex Dynamic Environment

LIU Boyi1, CHENG Jieren1,2+, TANG Xiangyan1, YIN Jianping3   

  1. 1. College of Information Science and Technology, Hainan University, Haikou 570228, China
    2. State Key Laboratory of Marine Resource Utilization in South China Sea, Hainan University, Haikou 570228, China
    3. State Key Laboratory of High Performance Computing, National University of Defense Technology, Changsha 410073, China
  • Online:2017-01-01 Published:2017-01-10

复杂动态环境下运动车辆的识别方法

刘博艺1,程杰仁1,2+,唐湘滟1,殷建平3   

  1. 1. 海南大学 信息科学技术学院,海口 570228
    2. 海南大学 南海海洋资源利用国家重点实验室,海口 570228
    3. 国防科学技术大学 高性能计算国家重点实验室,长沙 410073

Abstract: In view of the problem of obtaining the best threshold in vehicle recognition system, this paper presents a dynamic adaptive threshold method for vehicle recognition based on current research. This method firstly uses the image quality equalization algorithm based on entropy weight method to calculate the equalization of background image in traffic video. Secondly, the method obtains the best threshold of vehicle recognition by using?polynomial fitting. Thirdly, the method makes Gaussian fitting between the equalization of background image and the best threshold of   vehicle samples to get an adaptive threshold calculation model. The system updates the background image and calculates the equalization of background image real time. Then the system inputs the image equalization to the adaptive threshold calculation model and gets the best threshold of real-time vehicle recognition. The experimental results and theoretical analysis show that the method realizes the adaptive update of vehicle recognition according to the dynamic environment, and improves the accuracy of vehicle recognition in different environments.

Key words: vehicle recognition, image quality, Gaussian fitting, entropy weight method

摘要: 针对目前车辆识别方法在动态变化的复杂环境中车辆识别正确率低的问题,提出了一种基于动态自适应阈值的车辆识别方法。该方法首先利用基于熵权法的图像质量量化算法计算交通流视频中背景图像的质量值;然后通过对样本交通流设置的车辆检测阈值和基于该阈值识别车辆的正确率进行多项式拟合,获得该样本的车辆最佳检测阈值;最后对样本背景图像的质量值和样本车辆的最佳检测阈值进行高斯拟合,得到自适应阈值计算模型。该方法采用高斯混合模型实时获取交通流视频中的背景图像,计算背景图像的质量值,并输入到自适应阈值计算模型得到实时的车辆最佳检测阈值以识别车辆。实验和理论分析表明,该方法能根据动态变化的环境实时更新车辆检测阈值,有效地提高了车辆识别的正确率。

关键词: 车辆识别, 图像质量, 高斯拟合, 熵权法