Smoke Recognition Based on Multi-Scale Feature Fusion
Hidden dangers of fire are high in our country,seriously threatening the safety of production and life of people.Pollutants released by fire have potential effects on air quality and human health.Improving video fire smoke identification technology can effectively kill fire in the cradle.In this paper,a smoke recognition model based on multi-scale feature fusion is proposed.Based on neural network model VGG19,a top-down feature path and a bottom-up attention path are constructed to preserve low-level features while carrying out convolution,which is more conducive to the task of smoke recognition.In this study,the smoke image data set published by the State Key Laboratory of Fire Science was used for experiments,and the accuracy rate reached 99.75%,which was 9.85%higher than that of only using the neural network model VGG19.
万方数据
