摘 要：文章提出深度神经网络的方法，将分治策略引入到人流量统计问题中。这个方法以 VGG16 作为特征编码网络，UNet 作为解码网络，将人流量统计变成一个可拆解的任务。在统计图片上人数时，总可以把图片分解成多个子区域，使得每个区域的人流量计数都是在之前训练集上所见过的人数类别。然后再把每个子区域上面的人数加起来，就是统计结果。在对广州某个中国移动营业厅的人流量统计的实验中，预测的速度能达到实时监测的速度和高精确度。
关键词：分治策略；人流量统计；深度神经网络；VGG16；UNet
DOI:10.19850/j.cnki.2096-4706.2021.24.022
中图分类号：TP18 文献标识码：A 文章编号：2096-4706（2021）24-0084-05
Research on People Flow Statistics of China Mobile Business Hall Based on Deep Neural Network
CHEN Le, DING Ge, XIAO Zhongliang
(GMCC, Beijing 102200, China)
Abstract: In this paper, the method of deep neural network is proposed, and the divide and conquer strategy is introduced into the problem of people flow statistics. This method takes VGG16 as the feature coding network and UNet as decoding network, which turns the people flow statistics into a detachable task. When counting the number of people on the pictures, we can always decompose the picture into multiple sub-regions, so that the people flow count in each area is the number of people seen on the previous training set. Then add up the number of people in each sub-region, which is the result of statistics. In the experiment of people flow statistics in a China mobile business hall in Guangzhou, the predicted speed can reach real-time monitoring speed, and has high accuracy.
Keywords: divide and conquer strategy; people flow statistics; deep neural network; VGG16; UNe
参考文献：
[1] 中国移动 . 营运数据 [EB/OL].[2021-09-25].https:// www.chinamobileltd.com/sc/ir/operation_q.php.
[2] SAM D B，SAJJAN N N，BABU R V，et al. Divide and Grow：Capturing Huge Diversity in Crowd Images with Incrementally Growing CNN [C]//2018 IEEE/CVF Conference on Computer Vision and Pattern Recognition.IEEE：Salt Lake City，2018：3618-3626.
[3] BADRINARAYANAN V，KENDALL A，CIPOLLA R. SegNet：A Deep Convolutional Encoder-Decoder Architecture for Image Segmentation [J/OL].arXiv：1511.00561 [cs.CV].（2015-11-02）. https://arxiv.org/abs/1511.00561.
[4] HE K M，ZHANG X Y，REN S Q，et al. Delving Deep into Rectifiers：Surpassing Human-Level Performance on ImageNet Classification [J/OL].arXiv：1502.01852 [cs.CV].（2015-02-06）. https://arxiv.org/abs/1502.01852.
[5] SIMONYAN K，ZISSERMAN A. Very Deep Convolutional Networks for Large-Scale Image Recognition [J/OL].arXiv：1409.1556 [cs.CV].（2014-09-04）.https://arxiv.org/abs/1409.1556.
[6] LEMPITSKY C，ZISSERMAN A. Learning To count objects in images [C]//NIPS’10：Proceedings of the 23rd International Conference on Neural Information Processing Systems.Red Hook： Curran Associates，2010：1324-1332.
[7] LI R B，XIAN K，SHEN C H，et al. Deep attentionbased classification network for robust depth prediction [J/OL].arXiv： 1807.03959 [cs.CV].（2018-07-11）.https://arxiv.org/abs/1807.03959.
[8] NIU Z X，ZHOU M，WANG L，et al. Ordinal Regression with Multiple Output CNN for Age Estimation [C]//2016 IEEE Conference on Computer Vision and Pattern Recognition （CVPR）. Las Vegas：IEEE，2016：4920-4928.
[9] GIRSHICK R. Fast R-CNN [C]//2015 IEEE International Conference on Computer Vision （ICCV）.Santiago：IEEE，2015：1440-1448.
[10] IDREES H，TAYYAB M，ATHREY K，et al. Composition Loss for Counting，Density Map Estimation and Localization in Dense Crowds [J/OL].arXiv：1808.01050 [cs.CV].（2018-08-02）.https:// arxiv.org/abs/1808.01050.
[11] MASKA M，ULMAN V，SVOBODA D，et al. A benchmark for comparison of cell tracking algorithms [J]. Bioinformatics，2014，30（11）：1609-1617.
作者简介：陈乐（1982—），男，汉族，重庆人，项目总监， 硕士研究生，研究方向：AIOps、业务支撑系统运营支撑；丁戈 （1995—），男，汉族，广东广州人，项目经理，硕士研究生，研究方向：AIOps 算法；肖忠良（1986—），男，汉族，广东广州人， 项目经理，硕士研究生，研究方向：AIOps、业务支撑系统运营支撑。