Journal of Computer Applications ›› 2021, Vol. 41 ›› Issue (2): 544-549.DOI: 10.11772/j.issn.1001-9081.2020050623

Special Issue: 多媒体计算与计算机仿真

• Multimedia computing and computer simulation • Previous Articles     Next Articles

Dense crowd counting model based on spatial dimensional recurrent perception network

FU Qianhui, LI Qingkui, FU Jingnan, WANG Yu   

  1. School of Automation, Beijing Information Science and Technology University, Beijing 100192, China
  • Received:2020-05-12 Revised:2020-09-18 Online:2021-02-10 Published:2020-10-20
  • Supported by:
    This work is partially supported by the Promoting Connotative Development of University-Postgraduate Science and Technology Innovation Project (5121911048).


付倩慧, 李庆奎, 傅景楠, 王羽   

  1. 北京信息科技大学 自动化学院, 北京 100192
  • 通讯作者: 李庆奎
  • 作者简介:付倩慧(1996-),女,山东聊城人,硕士研究生,主要研究方向:图像处理、供应链系统;李庆奎(1971-),男,山东临沂人,教授,博士,主要研究方向:切换时滞系统、供应链系统;傅景楠(1993-),男,福建莆田人,硕士研究生,主要研究方向:图像处理、深度学习;王羽(1996-),女,北京人,硕士研究生,主要研究方向:图像处理、供应链系统。
  • 基金资助:

Abstract: Considering the limitations of the feature extraction of high-density crowd images with perspective distortion, a crowd counting model, named LMCNN, that combines Global Feature Perception Network (GFPNet) and Local Association Feature Perception Network (LAFPNet) was proposed. GFPNet was the backbone network of LMCNN, its output feature map was serialized and used as the input of LAFPNet. And the characteristic that the Recurrent Neural Network (RNN) senses the local association features on the time-series dimension was used to map the single spatial static feature to the feature space with local sequence association features, thus effectively reducing the impact of perspective distortion on crowd density estimation. To verify the effectiveness of the proposed model, experiments were conducted on Shanghaitech Part A and UCF_CC_50 datasets. The results show that compared to Atrous Convolutions Spatial Pyramid Network (ACSPNet), the Mean Absolute Error (MAE) of LMCNN was decreased by 18.7% and 20.3% at least, respectively, and the Mean Square Error (MSE) was decreased by 22.3% and 22.6% at least, respectively. The focus of LMCNN is the association between the front and back features on the spatial dimension, and by fully integrating the spatial dimension features and the sequence features in a single image, the crowd counting error caused by perspective distortion is reduced, and the number of people in dense areas can be more accurately predicted, thereby improving the regression accuracy of crowd density.

Key words: crowd counting, crowd density estimation, Convolutional Neural Network (CNN), Multi-column Convolutional Neural Network (MCNN), Long Short-Term Memory (LSTM) neural network

摘要: 考虑目前对具有透视畸变的高密度人群图像进行特征提取的局限性,提出了一种融合全局特征感知网络(GFPNet)和局部关联性特征感知网络(LAFPNet)的人群计数模型LMCNN。GFPNet是LMCNN的主干网络,将其输出的特征图进一步序列化并作为LAFPNet的输入,再利用循环神经网络(RNN)在时序维度上对局部关联性特征感知的特点将单一的空间静态特征映射到具有局部序列关联性特征的特征空间,从而有效地削减了透视畸变对人群密度估计造成的影响。为了验证所提模型的有效性,在Shanghaitech Part A子集和UCF_CC_50数据集上与原子卷积空间金字塔网络(ACSPNet)进行对比,结果表明所提模型的平均绝对误差(MAE)分别至少减小了18.7%和20.30%,均方误差(MSE)分别至少减小了22.3%和22.6%。LMCNN注重空间维度上前后特征的相关性,通过对空间维度特征与单图像内序列特征的充分融合,减小了由透视畸变引起的人群计数误差,能更加准确地预测密集区域人数,提高人群密度回归精度。

关键词: 人群计数, 人群密度估计, 卷积神经网络, 多列卷积神经网络, 长短时记忆神经网络

CLC Number: