Research on fire protection system based on deep learning

Flame detection systems are becoming more and more important in firefighting, and for indoor large space occasions, this paper designs a firefighting system based on computer vision. Combining the knowledge of mass kinematics and fluid dynamics, the trajectory of the firefighting water cannon jet is deduced, and the effect of different parameter conditions on the trajectory is analyzed using MATLAB simulation. The yolov4 flame detection model is trained to achieve accurate detection of the flame, and the coordinate value of the flame center position relative to the upper-left corner of the image is calculated at the same time. Then the horizontal turning angle required for the rotation of the fire gun is calculated according to the coordinate value, and the monocular visual localization technology is used to realize the localization of the fire source in the actual space, and the horizontal distance from the camera to the fire source is calculated. Substituting the horizontal distance into the jet trajectory algorithm can get the pitch angle of the fire cannon under the corresponding range.