A video surveillance system

Abstract

The present application belongs to the field of security technology, and in particular relates to a video surveillance system. The disadvantage of the current intelligent security video surveillance system is that data cannot be shared among various systems, so that each system can only use private data to learn and train models, and thus cannot fully utilize the advantages of big data to train high-quality models. At the same time, the rapid migration of models trained in different scenarios cannot be achieved, resulting in poor model generalization ability. The present application provides a video surveillance system, including a hardware perception layer, a data processing layer, a business logic layer, a cross-system-level networking layer, and a data storage unit that are sequentially connected through communication. For security monitoring, based on the federated transfer learning algorithm, in order to improve the learning ability and learning efficiency of the security monitoring system in the later stage, it solves the problem of low model training efficiency in the existing security monitoring system, user privacy data cannot be guaranteed, and cross-system data Unable to circulate, etc.

Description

technical field [0001] The present application belongs to the field of security technology, and in particular relates to a video monitoring system. Background technique [0002] The video monitoring system uses image monitoring as a means to monitor and record live images in real time. The video surveillance system allows security personnel to intuitively grasp the situation on the scene, and can analyze and collect evidence through video playback. The video surveillance system is an important part of the security system. The current video surveillance system is organically combined with the anti-theft alarm system to form a more reliable surveillance system. [0003] Google launched "Federal Learning" in 2016, and tens of millions of mobile phones jointly train a shared neural network model. This new method does not centralize the training data in one place, but uses multiple mobile phones scattered in the hands of millions of users to jointly train the machine learning m...

AI Technical Summary

Problems solved by technology

At the same time, the rapid migration of models trained in different scenarios cannot be realized, resulting in the problem of poor model generalization ability. This application provides a video surveillance system

Images