A Creative Research in the Field of Urban Fire Detection

As the most common disaster in human daily production and life, fire is very likely to threaten people’s life and property safety. In order to avoid fire, this paper designs and implements a fire detection system based on video and YOLOv5 model, and describes the improvement of YOLOv5 compared with the previous generation YOLO model. Collect fire images through the network, use image processing technology to enhance the data, make fire data set and divide the data set by setting aside method. Then the self-made fire data set is used to train the fire detection model based on YOLOv5 model, and the model is improved by adding transfer learning, multi-scale training and other technologies to improve the fire detection rate, and the corresponding weight files are obtained. Finally, the training method of the model and the ability of the model to detect fire are expounded, and the innovative application of the model in the field of fire detection is realized.