摘要
基于计算机视觉技术对道路交通视频中的运动目标进行分类。针对目标分割过程中的光线变换及分类效率问题,主要采用贝叶斯网络模型以及合适的前景提取模型以提高精度。提出一种改进的Vibe算法对运动目标进行检测,通过提取目标长宽比、Hu不变矩以及离散度特征等对目标进行分类。最终实验结果正确率在80%以上,说明该智能交通系统可以有效识别出运动目标,且具有较强的鲁棒性与适应性。
The purpose of this paper is to solve the problem of moving object classification in road traffic video which based on comput-er vision. Aiming at the problem of illumination change in target segmentation and efficiency of classification,this paper mainly uses Bayesian network model and appropriate foreground extraction model to improve the accuracy. First,an improved Vibe algorithm which can remove ghost is proposed to detect moving targets. The target is classified effectively by feature extraction of target aspect ratio ,Hu moment invariant,dispersion and the Bayesian classifier selection. The intelligent transportation system designed in this paper can ef-fectively identify the target,which has strong robustness and adaptability. The accuracy of the experiment is above 80%. Using Bayes-ian network model can achieve high real-time performance and better accuracy.
作者
周森鹏
穆平安
张仁杰
ZHOU Sen-peng;MU Ping-an;ZHANG Ren-jie(University of Shanghai for Science and Technology,School of Optical-Electrical and Computer-Engineering,Shanghai 200093,China)
出处
《软件导刊》
2019年第6期18-20,24,共4页
Software Guide
关键词
计算机视觉
智能交通
改进Vibe算法
目标检测
目标分类
computer vision
intelligent traffic system
improved Vibe algorithm
object detection
target classification