视频监控中运动车辆检测与跟踪算法研究
发布时间：2018-08-14 10:45
【摘要】：随着计算机技术的飞速发展,基于图像处理的智能交通系统因其实时、准确、高效的特点,受到了人们的广泛关注。在智能交通系统中,运动目标检测与跟踪可以完成对交通车辆的智能检测与跟踪、分类识别等功能。本文重点研究了智能交通中的运动车辆检测与跟踪算法。为了后期处理方便,提高对目标的识别跟踪效果,论文研究了常见的图像预处理方法,包括图像的复原、灰度化、二值化及形态学处理等。在对运动车辆检测算法进行分析研究的基础上,论文重点仿真了基于LucasKanade模型的光流法、基于Code Book算法的前景检测算法以及基于高斯模型的背景相减法等运动车辆检测算法;由于车辆阴影影响到检测效果,论文研究了车辆阴影的特点并进行了阴影消除;通过分析相关算法的优缺点,提出了结合基于边缘三帧差分法与混合高斯背景相减法的运动车辆提取算法。仿真实验表明,本算法可以完整的提取运动车辆区域,保留运动车辆的完整信息,提高了算法的实时性与鲁棒性。在对运动车辆跟踪算法进行分类研究后,论文以Kalman滤波和Mean Shift算法为基础,进行了车辆跟踪的算法仿真;在分析了交通车辆的特点后,用色调分量的概率分布建立特征空间,对Camshift算法进行了仿真研究;结合实际交通路况中车辆跟踪的常见问题,诸如车辆的遮挡,背景环境与目标相似,相似车辆毗邻等情况,仿真了将Camshift算法结合Kalman滤波的过程。
[Abstract]:With the rapid development of computer technology, the intelligent transportation system based on image processing has received extensive attention because of its real-time, accurate and efficient characteristics. In intelligent transportation system, moving target detection and tracking can accomplish the functions of intelligent detection and tracking, classification and recognition of traffic vehicles. This paper focuses on the moving vehicle detection and tracking algorithm in intelligent traffic. In order to facilitate the post-processing and improve the target recognition and tracking effect, this paper studies the common image preprocessing methods, including image restoration, grayscale, binarization and morphological processing. Based on the analysis and research of moving vehicle detection algorithm, this paper focuses on simulation of moving vehicle detection algorithms such as optical flow method based on LucasKanade model, foreground detection algorithm based on Code Book algorithm and background subtraction algorithm based on Gao Si model. Because of the influence of vehicle shadow on the detection effect, this paper studies the characteristics of vehicle shadow and eliminates the shadow, and analyzes the advantages and disadvantages of the related algorithms. A moving vehicle extraction algorithm based on edge three frame difference method and hybrid Gao Si background subtraction is proposed. The simulation results show that the proposed algorithm can extract the moving vehicle area completely and retain the complete information of the moving vehicle, and improve the real-time performance and robustness of the algorithm. After studying the classification of moving vehicle tracking algorithm, based on Kalman filter and Mean Shift algorithm, the simulation of vehicle tracking algorithm is carried out, and after analyzing the characteristics of traffic vehicle, the feature space is established by the probability distribution of hue component. The Camshift algorithm is simulated, and the process of combining the Camshift algorithm with Kalman filtering is simulated in combination with the common problems of vehicle tracking in actual traffic conditions, such as vehicle occlusion, similar background environment and target, similar vehicle proximity and so on.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN948.6;U495
[Abstract]:With the rapid development of computer technology, the intelligent transportation system based on image processing has received extensive attention because of its real-time, accurate and efficient characteristics. In intelligent transportation system, moving target detection and tracking can accomplish the functions of intelligent detection and tracking, classification and recognition of traffic vehicles. This paper focuses on the moving vehicle detection and tracking algorithm in intelligent traffic. In order to facilitate the post-processing and improve the target recognition and tracking effect, this paper studies the common image preprocessing methods, including image restoration, grayscale, binarization and morphological processing. Based on the analysis and research of moving vehicle detection algorithm, this paper focuses on simulation of moving vehicle detection algorithms such as optical flow method based on LucasKanade model, foreground detection algorithm based on Code Book algorithm and background subtraction algorithm based on Gao Si model. Because of the influence of vehicle shadow on the detection effect, this paper studies the characteristics of vehicle shadow and eliminates the shadow, and analyzes the advantages and disadvantages of the related algorithms. A moving vehicle extraction algorithm based on edge three frame difference method and hybrid Gao Si background subtraction is proposed. The simulation results show that the proposed algorithm can extract the moving vehicle area completely and retain the complete information of the moving vehicle, and improve the real-time performance and robustness of the algorithm. After studying the classification of moving vehicle tracking algorithm, based on Kalman filter and Mean Shift algorithm, the simulation of vehicle tracking algorithm is carried out, and after analyzing the characteristics of traffic vehicle, the feature space is established by the probability distribution of hue component. The Camshift algorithm is simulated, and the process of combining the Camshift algorithm with Kalman filtering is simulated in combination with the common problems of vehicle tracking in actual traffic conditions, such as vehicle occlusion, similar background environment and target, similar vehicle proximity and so on.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TN948.6;U495
【参考文献】
相关期刊论文 前1条
1 蔡s，
本文编号：2182609
本文链接：https://www.wllwen.com/kejilunwen/daoluqiaoliang/2182609.html
