Design of a home video behavior recognition system based on visual privacy security mechanism

The rapid development of the Internet and advanced technology has brought great convenience to people's lives; However, real-time video and other privacy information obtained from computers can be leaked, resulting in economic losses and not conducive to the construction of computer network security. In response to the above issues, this study introduces compressed perception theory and temporal adaptive modules to achieve visual shielding, and based on this, designs a home video behavior system based on visual privacy security mechanism. The research results show that in the comparison of measurement matrices at different levels, the Bernoulli random matrix has the highest recognition accuracy, with recognition accuracy rates of 100 %, 98.73 %, 98.76 %, and 85.62 % from the first layer to the fourth layer, respectively. In the recognition performance results of different video behavior recognition systems in the YouTube database, UCF Sports database, and Hollywood2 database, the average recognition accuracy of the proposed system is the highest in most cases, with 94.6 %, 73.5 %, and 77.1 %, respectively. In summary, the system proposed in the study can achieve accurate recognition of home video behavior after visual masking, and has good results in practical applications.