摘要
为了实现复杂场景中多运动行人目标的检测和跟踪,采用计算机视觉技术对真实场景下行人运动视频进行处理。利用基于高斯混合模型的背景消减法提取运动目标前景,并通过形态学运算进行目标清晰化处理。采用基于预测的Kalman滤波算法对运动目标进行跟踪,并对画面中的人数进行实时统计和更新。输出多运动行人目标的实时坐标和运动轨迹。研究结果表明:本文提出的算法能够快速准确地检测和追踪多运动行人目标,初步提取行人运动轨迹。
In order to achieve detection and tracking of multi-sport pedestrians in complex scenes,we processed the pedestrians'motion video in real scenes based on computer vision technology. First,the foreground of the moving object was extracted using the background subtraction method based on Gaussian mixture model. Morphological operation method was used to obtain clearer target. Then,Kalman filtering algorithm was applied to track the moving targets,calculating and updating the number of people in the screen.Finally,the real-time coordinates and motion trajectories of multi-motion pedestrians were output.The results indicate that the algorithm proposed in this paper can quickly and accurately detect and track multiple moving pedestrians,and extract pedestrian moving trajectories.
作者
谢玮
成艳英
陈柯
张玉春
XIE Wei;CHENG Yan-ying;CHEN ke;ZHANG Yu-chun(Faculty of Geosciences and Environmental Engineering, Southwest Jiaotong University,Sichuan Chengdu 610031,China)
出处
《消防科学与技术》
CAS
北大核心
2019年第1期44-47,共4页
Fire Science and Technology
基金
国家自然科学基金项目(51578464)
关键词
目标检测
目标跟踪
轨迹提取
图像处理
人员疏散
object detection
object tracking
trajectory extraction
Image processing
evacuation
