基于道路监控视频的交通拥堵判别方法研究
发布时间：2018-11-23 10:39
【摘要】：随着经济的快速发展,各个城市的汽车数量不断增加,道路的交通状况越发的复杂,实现交通状况的准确判别是解决道路拥堵问题的基础。道路监控系统的普及、图像处理与模式识别等技术的发展,使得基于视频的交通特征参数的提取实现交通状况的判别成为当前研究的热点。在实际的场景中道路信息系统的故障在所难免,容易造成道路交通流量数据的丢失,实现这些数据的修复显得尤为重要。为了解决道路状况判别这一问题,本文通过对道路交通视频的处理,获得道路交通特征参数,提出了一种基于核函数模糊C均值聚类(KFCM)的交通拥堵判别方法,同时将时空压缩感知压缩感知应用于道路交通流量数据的修复过程中。论文的主要工作如下:从实时交通视频中获得道路交通特征参数,首先要实现运动车辆的目标检测。本文对传统的像素级Vibe目标检测算法进行了改进,提出了一种基于阈值的自适应Vibe目标检测算法。针对检测中存在的鬼影,引入了基于Otsu阈值的鬼影抑制方法,将单个像素点的背景判别与整幅图像的特征相结合。为了更好地适应前景目标运动状况变化较大的情况,根据前景目标质心的运动速度,自适应的调整背景的更新速度。实验证明,本文的改进算法,能够快速有效的抑制鬼影,同时提高了目标检测的准确性和鲁棒性。其次,本文提出了一种基于KFCM的交通拥堵判别方法。交通拥堵的判别采用道路空间占道比、车流量以及道路宏观光流速度三个参数。对交通视频通过多帧融合进行道路的检测,计算前景目标像素个数与道路像素个数的比值获得道路空间占道;通过虚拟线圈法与Vibe算法结合统计车流量;融合了Harris角点检测算法以及H-S光流算法计算了整个车道的宏观光流速度。在此基础上,根据交通状态之间具有的模糊性,采用KFCM算法寻找交通状态的聚类中心,建立交通拥堵判别器,最后通过计算欧氏距离得到当前的交通拥堵状态。实验证明,本文提出的方法能够快速准确的进行道路拥堵状态的判别。最后,视频交通特征参数获取过程中交通流量参数可能丢失,道路交通流量的结构特性使其具有一定的冗余性和可压缩性,因此可将时空压缩感知理论应用于交通流量参数的修复中。本文构造了道路网络的交通流量矩阵,并结合道路流量的低秩性和时间-空间相关性的特点,提出了交通流量参数的时间相关矩阵和空间相关矩阵的构造方法,并利用近似矩阵对缺失元素进行插值重构实现流量数据的修复。该方法能够准确有效的修复缺失的交通流量参数。
[Abstract]:With the rapid development of economy, the number of cars in each city is increasing, and the traffic situation is becoming more and more complicated. The accurate identification of traffic condition is the basis of solving the problem of road congestion. With the popularization of road monitoring system and the development of image processing and pattern recognition technology, the extraction of traffic feature parameters based on video has become a hot research topic. In the actual scenario, the failure of road information system is inevitable, which can easily lead to the loss of road traffic flow data, so it is particularly important to realize the repair of these data. In order to solve the problem of road condition discrimination, the road traffic characteristic parameters are obtained by processing road traffic video, and a traffic congestion discrimination method based on kernel function fuzzy C-means clustering (KFCM) is proposed. At the same time, space-time compression perception is applied to the restoration of road traffic flow data. The main work of this paper is as follows: firstly, the target detection of moving vehicles should be realized by obtaining the characteristic parameters of road traffic from real-time traffic video. In this paper, the traditional pixel level Vibe target detection algorithm is improved, and an adaptive Vibe target detection algorithm based on threshold is proposed. In view of the existence of ghosts in the detection, a Otsu threshold based ghost image suppression method is introduced, which combines the background discrimination of a single pixel with the features of the whole image. In order to better adapt to the situation where the moving state of the foreground target changes greatly, the updating speed of the background is adjusted adaptively according to the velocity of the centroid of the foreground target. Experimental results show that the improved algorithm can suppress ghost images quickly and effectively, and improve the accuracy and robustness of target detection. Secondly, this paper proposes a traffic congestion discrimination method based on KFCM. Traffic congestion is judged by three parameters: road space ratio, vehicle flow rate and road macroscopic light flow speed. Traffic video is detected by multi-frame fusion, the ratio of foreground pixels to road pixels is calculated, and the traffic flow is calculated by virtual coil method and Vibe algorithm. Harris corner detection algorithm and H-S optical flow algorithm are combined to calculate the macro optical flow velocity of the whole driveway. On this basis, according to the fuzziness between traffic states, the KFCM algorithm is used to find the clustering center of traffic state, and the traffic congestion discriminator is established. Finally, the current traffic congestion state is obtained by calculating Euclidean distance. Experimental results show that the proposed method can quickly and accurately distinguish the traffic congestion. Finally, the traffic flow parameters may be lost in the process of obtaining video traffic characteristic parameters, and the structural characteristics of road traffic flow make it redundant and compressible. Therefore, the theory of space-time compression sensing can be applied to the restoration of traffic flow parameters. In this paper, the traffic flow matrix of road network is constructed, and combined with the characteristics of low rank and time-space correlation of road flow, the method of constructing time correlation matrix and spatial correlation matrix of traffic flow parameters is proposed. And the approximate matrix is used to interpolate and reconstruct the missing elements to repair the traffic data. This method can accurately and effectively repair the missing traffic flow parameters.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.265
本文编号：2351294
[Abstract]:With the rapid development of economy, the number of cars in each city is increasing, and the traffic situation is becoming more and more complicated. The accurate identification of traffic condition is the basis of solving the problem of road congestion. With the popularization of road monitoring system and the development of image processing and pattern recognition technology, the extraction of traffic feature parameters based on video has become a hot research topic. In the actual scenario, the failure of road information system is inevitable, which can easily lead to the loss of road traffic flow data, so it is particularly important to realize the repair of these data. In order to solve the problem of road condition discrimination, the road traffic characteristic parameters are obtained by processing road traffic video, and a traffic congestion discrimination method based on kernel function fuzzy C-means clustering (KFCM) is proposed. At the same time, space-time compression perception is applied to the restoration of road traffic flow data. The main work of this paper is as follows: firstly, the target detection of moving vehicles should be realized by obtaining the characteristic parameters of road traffic from real-time traffic video. In this paper, the traditional pixel level Vibe target detection algorithm is improved, and an adaptive Vibe target detection algorithm based on threshold is proposed. In view of the existence of ghosts in the detection, a Otsu threshold based ghost image suppression method is introduced, which combines the background discrimination of a single pixel with the features of the whole image. In order to better adapt to the situation where the moving state of the foreground target changes greatly, the updating speed of the background is adjusted adaptively according to the velocity of the centroid of the foreground target. Experimental results show that the improved algorithm can suppress ghost images quickly and effectively, and improve the accuracy and robustness of target detection. Secondly, this paper proposes a traffic congestion discrimination method based on KFCM. Traffic congestion is judged by three parameters: road space ratio, vehicle flow rate and road macroscopic light flow speed. Traffic video is detected by multi-frame fusion, the ratio of foreground pixels to road pixels is calculated, and the traffic flow is calculated by virtual coil method and Vibe algorithm. Harris corner detection algorithm and H-S optical flow algorithm are combined to calculate the macro optical flow velocity of the whole driveway. On this basis, according to the fuzziness between traffic states, the KFCM algorithm is used to find the clustering center of traffic state, and the traffic congestion discriminator is established. Finally, the current traffic congestion state is obtained by calculating Euclidean distance. Experimental results show that the proposed method can quickly and accurately distinguish the traffic congestion. Finally, the traffic flow parameters may be lost in the process of obtaining video traffic characteristic parameters, and the structural characteristics of road traffic flow make it redundant and compressible. Therefore, the theory of space-time compression sensing can be applied to the restoration of traffic flow parameters. In this paper, the traffic flow matrix of road network is constructed, and combined with the characteristics of low rank and time-space correlation of road flow, the method of constructing time correlation matrix and spatial correlation matrix of traffic flow parameters is proposed. And the approximate matrix is used to interpolate and reconstruct the missing elements to repair the traffic data. This method can accurately and effectively repair the missing traffic flow parameters.
【学位授予单位】：南京邮电大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：U491.265
【参考文献】
相关期刊论文 前1条
1 徐健锐;李星毅;施化吉;;处理缺失数据的短时交通流预测模型[J];计算机应用;2010年04期
，本文编号：2351294
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2351294.html
