基于视频图像的复杂背景中流动人群数目的识别

Identifying the flowing passenger numbers in complex environments by video images

客流数据的实时采集是轨道交通系统中客流安全疏散研究的关键问题之一,利用图像处理方法来进行流动人群监测在国内外已有研究,但在人流动性大并且数量多的复杂场景的监测并不令人满意。本文利用普通CCD采集图像,综合帧间图像差异和消除背景图像的处理技术来有效检测图像中的运动人群,克服传统方法中存在的测量误差,并依据所检测的前景图像区域面积与整个图像有效面积的比例来估测当前图像中人群数目。实验数据显示该方法在给定误差内能获得较好的结果。

Real time collecting of passenger flow data is one of the key issued on safely evacuating passenger flows in the railway traffic system. There have been researches inside and outside on monitoring flows by using image processing methods , however, those methods donot prove in complex environments with large and highly moving crowds. In our study, we use CCD to get images, integrate frame differences and abstract background images to detect passenger flows, which overcomes the problem of errors occurring in traditional measure ments; we estimate the passenger numbers of the crowds by calculating the ratio of the foreground image area to the total available image area. The experimental data indicate that the method works well within the range of given errors.