1. Project Introduction Fire is a common and harmful disaster in our daily life. It is easy to cause huge property losses, destroy the ecological balance, cause adverse social and political effects, and most importantly, harm people's lives. The current fire extinguishing methods are mainly fire extinguishers, sprinklers, fire alarms, etc. These fire extinguishing methods also have shortcomings such as fire extinguishing substances polluting the environment, consuming huge manpower, and missing the best time for fire extinguishing. With the advancement of science and technology, the addition of artificial intelligence has brought about earth-shaking changes in people's lives. The combination of science and life has given our team a new idea for fire-intelligent fire protection. We believe that imperfect fire extinguishing devices in daily life give fire a great opportunity. Therefore, a set of perfect fire extinguishing device is the top priority of fire fighting work. The team developed a fire-extinguishing device (ML-based thermal imaging high-pressure spray fire extinguishing device), mobilized AI chips to work through ML, established data sets to simulate scenes and deal with problems, developed a set of systems focusing on supporting monitoring, analysis, adjustment and decision-making, and combined big data network and database through analysis and processing of image data returned by high-performance thermal imaging instruments, the scientific fire extinguishing mode of timely detection, rapid response and accurate fire extinguishing is realized. Technical Introduction of 2. Products The ML is to build mathematical models based on sample data through algorithms, collect all kinds of data from simulation scenarios to build dense data sets to build training models, and conduct large amounts of data training to guide computers to learn, so as to achieve the purpose of prediction and decision-making without explicit programming to perform tasks. It belongs to a subset of artificial intelligence. After a long period of study and research, the team has a basic grasp of ML. At present, the team is trying to apply the ML to the monitoring module of the device, the team technicians use the ML supervision learning method to process the data, the use of feature learning algorithm and image recognition, detection, cutting, clustering, noise reduction and other processing, for the material, ignition point, ignition mode, ignition speed, flame style, color temperature value and heat distribution of all kinds of fire substances, all the features in the whole process of fire are classified to establish mathematical models and then gathered to establish dense data sets. MATLAB and Python are used to establish neural networks. On this basis, a large number of logical units are established to deal with necessary and unexpected situations, then the data model training through regression algorithm, instance-based algorithm and other algorithms matching trial, so that the instrument to summarize the fire under various circumstances, the establishment of scientific linear regression results to achieve no clear coding of the fire in a timely manner to predict and make decisions. The tripod head is a device that plays the role of balance and stability. The combination of the tripod head and thermal imaging instrument can effectively improve the stability and accuracy of the whole device. The tripod head we adopt is an indoor hoisting spherical tripod head. The advantages of this tripod head are that it can operate with a horizontal angle of 360 ° omnidirectional cruise and a vertical angle of -20 ° (depression angle)-+90 ° (elevation angle), the PTZ control method adopted by the technicians of this team is to use SDK development kit to use JAVA language to program and realize the communication between images, AI chips and computers by connecting the camera support package provided in MATLAB. Through the preparation of JAVA PTZ control instruction function, the control of the device performance such as the movement direction, angle, speed, preset position, cruise scan, power-on initial action of the steering gear is realized. In view of the suddenness of the fire and the smallness of the previous flame, we have adopted a spherical dual-spectral thermal imaging camera with resolutions of 384x 288 and 640x 480 respectively and a thermal sensitive sensing range of 50mk @ 25. The frame rate of the lens is 50Hz and 30Hz, the response band is between 8 and 14μm, and the temperature measurement range is -20°C ~ 150°C/0°C ~ 300°C/60°C ~ 600°C. The temperature measurement method adopts the mode of combining point temperature measurement and area temperature measurement, automatic switching, electric focusing and automatic focusing functions, and digital zoom is 2 times and 4 times, the image resolution is 1920x 1080(PAL system), the image resolution is 2 million pixels, the lens adopts 20 times optical lens, the horizontal viewing angle can reach 61.4 to 2.9, and the supplementary light adopts double LED supplementary light, which can monitor a circular area with a radius of about 10 meters. At the same time, we innovatively combine the mini booster pump and jet nozzle into the thermal imaging instrument through scientific spatial distribution. 3. Competition Analysis Our products are benchmarked against traditional fire sprinkler equipment. At present, many researchers are committed to the exploration of fire sprinkler equipment, its working principle is through the fire sprinkler head on the red liquid on the temperature sensing to achieve. When the temperature rises, it expands rapidly, breaking the glass in which it is housed, and then the pressure sensor inside the glass causes the fire sprinkler pump to spray water. It is also used in different occasions. For example, ordinary sprinklers are installed on the spray pipe network, spraying 40%-60% of the total water volume downward, and spraying the larger part to the ceiling. More for restaurants, shops, warehouses, underground garages and other places. Side wall type sprinkler installed against the wall, more used in the space pipe is difficult to install, mainly used in the office, lobby, lounge, corridor, living room. The cover of the concealed spray is welded to the thread with fusible metal, and the melting point is 57 degrees. Therefore, in the event of a fire, the temperature rises first to cause the lid to fall off, and when the temperature rises to 68 degrees (general sprinkler head), the glass tube bursts and the water flows out. Concealed spray is mostly used in high-end hotels, residences, theaters and other places that need to ensure the smooth and clean effect of smallpox. However, the most taboo of concealed nozzles is that the cover is stained with paint and paint, which will cause the action to fail. But the above several kinds of spray device there are some security risks. First of all, the lowest melting point triggered by the spray device of these spray devices is 68 °, which will cause it to have no response to subtle flames and have a high trigger threshold; secondly, due to its design limitations, when it is triggered, it will lead to Large-scale spraying does not have good pertinence, and it will also cause greater loss of manpower and material resources; and, when it is started, the sprayed fire extinguishing substance is only water, and different fire causes require different fire extinguishing substances. Even if it is started, the accuracy of the fire extinguishing method cannot be guaranteed. The ML-based thermal imaging high-pressure jet fire extinguishing device developed by our team has the greatest advantages of precise positioning, intelligent deployment of fire extinguishing materials and effective extinguishing. Our products combine thermal imaging and ML technology. We first use thermal imaging technology to achieve accurate positioning of indoor fire sources, and then process data through chips with ML functions, quickly match with the database, and use appropriate fire extinguishing materials to more accurately and effectively extinguish flames, which greatly reduces errors in the fire extinguishing process and reduces the risk of fire.