车辆安全例检的图像识别与跟踪定位方法研究-《集美大学学报（自然科学版）》

文章信息/Info

Title:: Research on Image Recognition and Tracking Method of Vehicle Safety Inspection

作者:: 郭宝英1; 郑思凡2; 3; （1.仰恩大学工程技术学院，福建泉州 362000；2.黎明大学机电工程与自动化学院，福建泉州 362000；3.福州大学物理与信息工程学院，福建福州 350116）

Author(s):: GUO Baoying1; ZHENG Sifan2; 3; (1.College of Engineering,University of Yan’en,Quanzhou 362000,China;2.College of Mechanical and Electrical Engineering and Automation,Liming University,Quanzhou 362000,China;3.College of Physics and Information Engineering,Fuzhou University，Fuzhou 350116,China)

Keywords:: motion history image; motion detection; target tracking; vehicle safety inspection; image recognition

摘要:: 为满足车辆安全例检智能化和信息化的需要，研究了室外复杂场景下例检车辆的图像识别与跟踪定位问题。针对运动历史图算法在运动检测中出现目标轮廓的拖尾、空洞、边缘不连续等缺点，引入了高斯塔式分解及形态学滤波方法对轮廓进行补偿修正。针对跟踪过程中镜头抖动及积水倒影造成的多目标干扰，提出了一种缩小目标检测区域与图像帧数并结合车体颜色直方图和车牌SIFT描述子逐帧匹配的跟踪方法。实验结果证明，在不损失鲁棒性与准确性的情况下，完成了车辆的实时跟踪。

Abstract:: For the needs of informationization and automation of vehicle safety inspection,the problem of vehicle recognition and tracking under outdoor complex scene is studied.In order to solve the problems of motion history image algorithm such as tailing,voiding and edge discontinuity in motion detection,Gaussian decomposition and morphological filtering are used to compensate the contours.In order to eliminate the multitarget interference caused by lens jitter and water reflection during the tracking,a method is proposed to reduce the target detection area and image frame number.The new method performs frame by framematching via the vehicle color histogram and license plate SIFT description.Experiments show that this study can realize a realtime tracking of the vehicle without loss of robustness and accuracy,and has good commercial application prospect.