Smoke video detection based on lightweight convolutional neural network

In this paper, a novel method is used to detect smoke from video sequences, which combines the traditional smoke detection algorithm with the current popular lightweight convolutional neural network. The method of combining artificial smoke feature extraction with neural network automatic smoke feature extraction is adopted. The lightweight neural network MobileNet model is reconstructed and trained to solve the problem of smoke detection classification. Compared with the current popular smoke detection algorithm, this method had better real- time performance, improved the smoke detection accuracy effectively and reduced the false alarm rate of smoke detection.