基于MCNN人群密度估算的安全预警

Safety Warning Based on MCNN Crowd Density Estimation

针对大规模人群聚集造成的拥挤和踩踏事件给城市公共安全带来的巨大损失,确定高密度人群区域进行安全预警变得至关重要。使用高斯滤降低噪声,输入到多列卷积神经网络MCNN(Multi-column Convolutional Neural Network),将图像映射为人群密度图,准确的估算人群数量,确定出高密度人群区域。在Shanghaitech数据集上实验,test_data_A部分MSE和MAE分别为229.55和162.58,test_data_B部分MSE和MAE分别为43.68和25.61。

In view of the huge loss of urban public safety caused by crowd and stampede caused by largescale crowd gathering,it is crucial to determine the high-density crowd area for safety early warning.The Gaussian filter is used to reduce the noise,and the image is input into the Multi-column Convolutional Neural Network(MCNN),and the image is mapped into the crowd density map,the number of people is accurately estimated,and the high-density crowd area is determined.Experiments on the Shanghaitech dataset show that the MSE and MAE of the test_data_A part are 229.55 and 162.58,and the MSE and MAE of the test_data_B part are 43.68 and 25.61,respectively.