Journal of Computer Applications ›› 2024, Vol. 44 ›› Issue (7): 2243-2249.DOI: 10.11772/j.issn.1001-9081.2023060782

• Multimedia computing and computer simulation • Previous Articles     Next Articles

Crowd counting algorithm with multi-scale fusion based on normal inverse Gamma distribution

Wei LI1(), Xiaorong ZHANG1, Peng CHEN1, Qing LI2, Changqing ZHANG2   

  1. 1.The 28th Research Institute,China Electronics Technology Group Corporation,Nanjing Jiangsu 210007,China
    2.College of Intelligence and Computing,Tianjin University,Tianjin 300354,China
  • Received:2023-06-27 Revised:2023-08-24 Accepted:2023-08-25 Online:2023-09-04 Published:2024-07-10
  • Contact: Wei LI
  • About author:ZHANG Xiaorong, born in 1989, M. S., engineer. Her research interests include argumentation on the field of armed police command and control.
    CHEN Peng, born in 1978, senior engineer. His research interests include development planning and overall argumentation in the field of security (such as armed police).
    LI Qing, born in 1995, Ph. D. candidate. His research interests include computer vision.
    ZHANG Changqing, born in 1982, Ph. D., associate professor. His research interests include machine learning, computer vision.
    First author contact:LI Wei, born in 1990, M. S., engineer. His research interests include application in the field of armed police command and control.
  • Supported by:
    National Natural Science Foundation of China(61976151)

基于正态逆伽马分布的多尺度融合人群计数算法

李伟1(), 张晓蓉1, 陈鹏1, 李清2, 张长青2   

  1. 1.中国电子科技集团公司 第二十八研究所, 南京 210007
    2.天津大学 智能与计算学部, 天津 300354
  • 通讯作者: 李伟
  • 作者简介:张晓蓉(1989—),女,山西太原人,工程师,硕士,主要研究方向:武警指挥和控制领域论证;
    陈鹏(1978—),男,山东蒙阴人,正高级工程师,主要研究方向:安全(含武警)领域发展规划和总体论证;
    李清(1995—),男,天津人,博士研究生,主要研究方向:计算机视觉;
    张长青(1982—),男,河南安阳人,副教授,博士,CCF会员,主要研究方向:机器学习、计算机视觉。
    第一联系人:李伟(1990—),男,吉林吉林人,工程师,硕士,主要研究方向:武警指挥和控制领域应用;
  • 基金资助:
    国家自然科学基金资助项目(61976151)

Abstract:

To solve the problem of large variation caused by different distances between monitoring camera and crowd in the crowd analysis tasks, a crowd counting algorithm with multi-scale fusion based on normal inverse Gamma distribution was proposed, named MSF (Multi-Scale Fusion crowd counting) algorithm. Firstly, the common features were extracted with the traditional backbone, and then the pedestrian information of different scales was obtained with the multi-scale information extraction module. Secondly, a crowd density estimation module and an uncertainty estimation module for evaluating the reliability of the prediction results of each scale were contained in each scale network. Finally, more accurate density regression results were obtained by dynamically fusing the multi-scale prediction results based on the reliability in the multi-scale prediction fusion module. The experimental results show that after the expansion of the existing method Converged Scene Recognition Network (CSRNet) by multi-scale trusted fusion, the Mean Absolute Error (MAE) and Mean Squared Error (MSE) of crowd counting on UCF-QNRF dataset are significantly decreased by 4.43% and 1.37%, respectively, which verifies the rationality and effectiveness of MSF algorithm. In addition, different from the existing methods, the MSF algorithm can not only predict the crowd density, but also provide the reliability of the prediction during the deployment stage, so that the inaccurate areas predicted by the algorithm can be timely warned in practical applications, reducing the wrong prediction risks in subsequent analysis tasks.

Key words: crowd counting, multi-scale, trustworthy fusion, crowd density estimation, uncertainty

摘要:

针对人群分析任务中往往存在的因监控与人群距离不同而导致的尺度变化大的问题,提出一种基于正态逆伽马分布的多尺度融合人群计数算法MSF(Multi-Scale Fusion crowd counting)。首先,使用传统骨架提取公共特征,通过多尺度信息提取模块获得图像中不同尺度的行人信息;其次,每个尺度的网络各自包含一个人群密度估计模块和一个用于评估每个尺度预测结果可信度的不确定估计模块;最后,多尺度预测融合模块依据可信度对多尺度预测结果进行动态融合,以获得更准确的密度回归结果。实验结果表明,现有算法密集场景识别网络(CSRNet)在通过多尺度可信融合扩展后,在UCF-QNRF数据集上人群计数的平均绝对误差(MAE)和均方误差(MSE)分别减小了4.43%和1.37%,验证了MSF算法的合理性和有效性。此外,与现有算法不同,MSF算法不仅可以预测人群密度,还可以在部署阶段提供预测的可信程度,从而使算法在实际应用中能及时预警模型预测不准确的区域,降低后续分析任务出现错误预判的风险。

关键词: 人群计数, 多尺度, 可信融合, 人群密度估计, 不确定性

CLC Number: