This project is based on the design of existing video image capture and input equipment, combined with cutting-edge technologies such as artificial intelligence algorithms, intelligent control, computer vision, etc., to embed deep learning algorithms into surveillance cameras to develop an intelligent and efficient distributed intelligent video image fire warning system. In the current new form, the focus of fire prevention is more on the early warning of fire rather than the rescue and follow-up resettlement of disaster outbreak. However, the fire detection and alarm products that are widely used are mostly based on sensor principle technology. In these technical methods, a characteristic physical quantity in the process of fire is taken as the monitoring object, and in some special occasions, it can not send out effective fire alarm signal or even fail in the face of large open field, and there are some defects in real-time. The existing fire detection based on image recognition technology also has problems such as insufficient small target analysis, inaccurate accuracy, and image analysis speed needs to be improved. In summary, we designed and developed a real-time fire detection and early warning system based on deep learning YOLO target detection algorithm. The system takes the camera as the carrier, carries the video image fire detection software, is equipped with video coding equipment, network switching equipment, etc. and is assisted by the traditional smoke sensor. Through the deep learning algorithm, the convolutional neural network is used to extract the features of the digital image, and then the classifier is used to identify the smoke and flame. This fire detection system based on target detection technology not only improves the accuracy of target recognition, but also responds faster. It can detect flames and smoke at a speed of 60FPS per second, and can respond to the alarm sound within 20ms. The applicable fields are also more diverse, which can effectively adapt to complex scenes and large open fields. The project, combined with the current hot spot "smart fire fighting", can also transmit video images to the main control room in real time, and can also upload data to the cloud through WiFi or 5G module, which can be viewed in real time through relevant APP and WeChat applets. In general, this project-distributed intelligent video image fire warning system based on target detection algorithm takes precedence over the current fire detection and alarm system in terms of algorithm and function. The accuracy and real-time grasp of the disaster situation are greatly improved. At the same time, the project also has the function of information data statistics, which can accumulate records of fire information and realize data traceability.