Abstract:

In order to improve the fire management of transformer substation, the video and image flame feature extraction and identification are achieved. The stream data of video surveillance is accessed. The flame identification model based on convolution neural network is built. The flame features are real time identified, and early warning and forecast are performed. Experiments showed that, the method can extract the flame feature automatically, and efficiently improve the accuracy of flame identification with complex background, with good robustness and generalization ability, and have great application prospect in the fire management of transformer substation.  

Key words:

convolution neural network, flame identification, transformer substation, fire management