Abstract: Aiming at the problems of prominent fire safety risks, difficult manual inspection and low emergency response efficiency, this paper designs an intelligent inspection and fire-fighting vehicle. The vehicle takes the STM32 microcontroller as the core, and adopts a closed-loop working mode of fixed-line tracking, fire source detection and disposal, and automatic return to tracking. Stable and accurate driving along the preset path is realized by grayscale tracking technology combined with cascade proportional-integral-derivative (PID) algorithm. A multi-source heterogeneous fire detection system is constructed by flame sensors, comprehensive environmental sensors and YOLOv11 visual recognition, and accurate fire detection and positioning are realized with the Kalman filtering algorithm. Experimental results show that the system stably completes the operations of path tracking, fire source detection and fire extinguishing, and realizes autonomous inspection in a simulated scenario. The system achieves a fire source recognition accuracy of 96.3% and an average fire extinguishing time of less than 6 s, which rapidly handles incipient fires and presents good engineering application value.

Key words: multi-source heterogeneous data; automatic inspection; fire-fighting vehicle; fire detection

摘要： 针对火灾安全风险突出、人工巡检难度大、应急处置效率低等问题，文章设计一款智 能巡检灭火小车。小车以 STM32 微控制器为核心，采用固定线路巡线、火源检测处置、自动 回归巡线的闭环工作模式。利用灰度循迹技术结合串级比例-微分-积分（PID）算法，实现小 车沿预设路径稳定精准行驶。通过火焰传感器、综合环境传感器与 YOLOv11 视觉识别构建多 源异构火情检测系统，配合卡尔曼滤波算法，完成火情精准探测与定位。实验结果表明，该 系统可稳定完成巡线、火源检测、灭火等作业流程，在模拟场景下实现自主巡检，火源识别 准确率达 96.3%，平均灭火时间小于 6 s，能够快速处置初期火灾，具备较好的工程应用价值。

关键词: 多源异构数据；自动巡检；灭火小车；火情检测