Design of Embedded Real-time Video Fire AI Recognition System

In order to solve the problem of the current video flame detection algorithm, such as low accuracy, high false alarm rate and high computing power requirements of the hardware system, a cost-effective embedded fire AI recognition system was proposed by referring to the idea of deep neural network. The system transplanted the lightweight algorithm MobileNet-SSD, and combined the multi-feature fusion technology to improve the small target detection ability of the algorithm. The experimental results show that this system overcomes the shortcomings of traditional fire detection technology, not only improves the accuracy of flame detection, but also reduces hardware costs, so it has greater practical value and market prospects.