Research on fall behavior detection algorithm

• 1、School of Information Engineering, Engineering University of PAP, Xi’an 710086, China

Abstract：In order to realize the real-time detection of falling behavior in the crowd and prevent the occurrence of stampede events, an improved YOLOv5s fall behavior detection algorithm is proposed in view of the problems of real-time detection of falling behavior and insufficient feature extraction ability. The network detection accuracy is enhanced and the amount of computation is ensured by rational combination of improving the basic residual block, adding a mixed domain attention mechanism to the backbone network and introducing a bidirectional feature pyramid structure to the neck. The experimental results show that, compared with the original network, the detection accuracy of the proposed algorithm is improved from 94.1% to 97.0%, the accuracy value is increased from 91.2% to 95.4%, the optimized algorithm detection speed can reach 0.028 s, and the number of frames per second can reach 36.3, which meets the real-time requirements.

Keywords： deep learning YOLOv5 fall behavior detection feature pyramid convolutional attention mechanism