复杂动态环境下运动车辆的识别方法

发布时间：2018-02-25 15:45

本文关键词： 车辆识别图像质量高斯拟合熵权法　出处：《计算机科学与探索》2017年01期 　论文类型：期刊论文

【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。
[Abstract]:In view of the problem that the current vehicle recognition methods have low accuracy in the dynamic and complex environment, A vehicle recognition method based on dynamic adaptive threshold is proposed. Firstly, the image quality quantization algorithm based on entropy weight is used to calculate the quality of background image in traffic flow video. Then by polynomial fitting the vehicle detection threshold set by the sample traffic flow and the correct rate of the vehicle recognition based on the threshold, the optimal detection threshold of the vehicle is obtained. Finally, Gao Si fitting the quality value of the sample background image and the optimal detection threshold of the sample vehicle, the adaptive threshold calculation model is obtained, and the background image in the traffic flow video is obtained in real time by the Gao Si mixed model. The quality value of the background image is calculated, and the real-time optimal vehicle detection threshold is obtained by inputting the adaptive threshold calculation model to identify the vehicle. The experimental and theoretical analysis shows that the method can update the vehicle detection threshold in real time according to the dynamic changing environment. The accuracy of vehicle recognition is improved effectively.
【作者单位】：海南大学信息科学技术学院;海南大学南海海洋资源利用国家重点实验室;国防科学技术大学高性能计算国家重点实验室;
【基金】：国家自然科学基金海南省自然科学基金海南大学博士启动基金;海南大学青年基金~~
【分类号】：TP391.41

【相似文献】