基于C3D-GRNN模型的人群异常行为识别算法

针对传统人群行为识别算法受人工主观因素影响较大等问题，综合三维卷积神经网络（C3D）与广义回归神经网络（GRNN）的优势和特点，提出并实现了基于C3D-GRNN模型的人群异常行为识别算法。该算法首先采用ViBe算法确定运动目标区域，然后通过改进C3D网络提取目标的HOG时空特征，再加入GRNN层进行分类训练，最后根据训练好的C3D-GRNN模型完成人群异常行为的识别与分类。实验结果表明：所提算法的HOG时空特征能够明显提升对人群行为的表达能力，减少了特征提取工作量，并且该方法的准确度和鲁棒性均高于支持向量机等其他同类方法，为小样本数据集的分类问题提供解决新思路，具有较高的应用价值。

An Algorithm for Identifying Crowd Abnormal Behavior Based on C3D-GRNN Model

In order to solve the problem that the traditional crowd behavior recognition algorithm is greatly affected by subjective factors,an algorithm for identifying crowd abnormal behavior based on C3D-GRNN model is proposed and implemented by synthesizing the advantages and characteristics of C3D and GRNN.Firstly,the ViBe algorithm was used to determine the moving target area.Then the HOG spatio-temporal features of the target were extracted by improving the C3D network.And then the GRNN layer was added for classification training.Finally,the identification and classification of the crowd abnormal behavior were completed according to the trained C3D-GRNN model.The experimental results show that the HOG spatio-temporal features of the proposed algorithm can significantly improve the ability to express crowd behavior and reduce the workload of feature extraction,and its robustness and accuracy are higher than other similar methods such as SVM.It also provides a solution for the classification problem of small sample data sets,and has high application value.