Automatic Directional Firefighting Water Cannon Control System

Integrating YOLO v5 for flame pixel coordinate acquisition, this paper analyzes the principles of binocular camera ranging and the conversion relationship between pixel and camera coordinates. A system is designed to calculate the true three-dimensional coordinates of flames relative to the camera using flame pixel coordinates and the camera's intrinsic matrix, facilitating flame recognition and localization. Additionally, Deeplab v3+ is employed in a side-view camera to segment flame jets and fit curves, measuring the angular deviation between flames and jets. Utilizing the PID control algorithm to steer the nozzle in an intelligent firefighting cannon control system. Through multiple experiments, the system's accuracy and stability are verified, demonstrating its reliability in real-world scenarios.