Design of Abnormal Behavior Detection System Based on Multimodal Fusion
Xiaoting Niu, Guanghai Zheng · Procedia Computer Science · 2024
The application of intelligent video surveillance systems can reduce the workload of staff, save human resources, and timely handle abnormal situations, reducing losses. Abnormal behavior detection based on multimodal fusion is a key part of intelligent video surveillance systems and a hot research topic in the field of pattern recognition. Its application prospects are very broad. Firstly, the dual stream residual convolutional neural network was studied. Based on the convolutional neural network model, the dual stream convolutional neural network model and residual convolutional neural network model were studied; Then, research the overall solution and related technologies, based on the multimodal fusion process, and study techniques such as RGB images, optical flow images, and bone images; Finally, the Resnet34 model was used for training and simulation, and the simulation results showed that the Resnet34 model has superiority. Applying deep residual structures to dual stream CNN has better recognition performance.