一种基于智能视频分析的人流量统计算法

Pedestrian Volume Counting Algorithm Based on Intelligent Video Analysis

针对安全出口处人流量统计实时性差、准确率低的问题,提出一种基于智能视频分析的人流量统计算法。首先,分别提取训练样本的梯度范数特征和梯度方向直方图特征,训练用于生成候选区域的分类器1和用于行人分类与定位的分类器2,其中分类器2在候选区域分类并定位行人,完成行人检测;再用Camshift跟踪算法跟踪检测到行人,并通过基于欧式距离和巴氏系数的数据关联算法建立当前帧的行人轨迹;最后分析轨迹,判断行人是否越过计数线,完成行人双向计数。实验结果表明,与基于传统梯度方向直方图特征和滑动窗口法的算法相比,该算法实时性较好,准确率有所提升,可有效完成人流量统计。

To address the low accuracy and poor real-time capability of pedestrian volume counting at fire exit,a pedestrian volume counting algorithm based on intelligent video analysis is proposed. Firstly, normed gradients feature and histograms of oriented gradients feature are extracted to train classifierⅠand classifierⅡ, respectively; the region proposal is generated with classifierⅠ, while in the region proposal classifierⅡis used to classify and locate pedestrian, thus pedestrians are detected successfully. Then, the detected pedestrians are tracked with Camshift algorithm, and pedestrian trajectories of current frame are established with data association algorithm based on Euclidean distance and Bhattacharyya coefficient. Finally, trajectories are analyzed and bidirectional pedestrian volume counting is realized. The experiment results show that the statistical accuracy and real-timecapability of the proposed algorithm is improved, compared to the algorithm based on traditional histograms of oriented gradients feature and sliding window method,so pedestrian volume counting is realized effectively.