摘要
针对密集场景下因尺度变化和遮挡等现象导致的人群计数精确率不高,在HRNet中将真值图生成为互不重叠的独立区域,便于网络密度图人群位置统计;然后引入3D注意力机制,使得网络专注于特征图的有用信息;最后在训练时将均方误差损失(MSE loss)、L1损失和交叉熵损失(Cross Entropy loss)相结合作为总的损失函数,优化模型泛化能力.上述方法的结合提高了模型在人群计数及人群位置定位中的准确性.将该模型在公开数据集NWPU、Shanghai Tech和UCF-QNRF中与近年来的主要方法进行对比,实验结果表明该模型可有效提升人群位置计数问题的准确性和鲁棒性.
Aiming at the low accuracy of crowd counting caused by scale change and occlusion in dense scenes,this paper proposes to generate the truth map into non-overlapped independent areas in HRNet to facilitate the crowd location statistics of network density map;Then the 3D attention mechanism is introduced to make the network focus on the useful information of the feature map;Finally,during the training,the mean square error loss(MSE loss),L1 loss and cross entropy loss are combined into the total loss function to optimize the generalization ability of the model.The combination of the above methods improves the accuracy of the model in crowd counting and crowd location.Compared with the main methods in recent years in the public data sets of NWPU,Shanghai Tech and UCF-QNRF,the experimental results show that the proposed model can effectively improve the accuracy and robustness of crowd location and counting.
作者
张天飞
龙海燕
丁娇
周荣强
ZHANG Tianfei;LONG Haiyan;DING Jiao;ZHOU Rongqiang(School of Electrical and Electronic Engineering, Anhui Institute of Information Technology, Wuhu, Anhui 241000, China;Hangzhou Zhiling Technology Co. Ltd, Hangzhou, Zhejiang 310000, China)
出处
《平顶山学院学报》
2022年第2期44-49,共6页
Journal of Pingdingshan University
基金
安徽省高校自然科学研究重点项目(KJ2020A0822)。
