Study on Fire Identification Efficiency of UAV Thermal Infrared Technology

Focusing on a subtropical Pinus massoniana forest, a dual-gradient experiment with canopy closure ranging from 0 to 1 and flight altitude from 20 to 140 m was conducted to develop a ternary coupling model of canopy closure, flight altitude, and first response time. Response surface analysis and random forest were employed to systematically quantify the synergistic effects and nonlinear patterns between flight parameters and canopy structure. Results indicated that first response time significantly prolonged with increases in flight altitude and canopy closure, and a strong positive interaction effect existed between the two factors (P < 0.001). When flight altitude exceeded 80 m and canopy closure was greater than 0.6, recognition performance deteriorated sharply, showing a “threshold abrupt change” feature. In the range of 50–80 m altitude and 0–0.6 canopy closure, first response time could be controlled within 10 s, representing the optimal performance zone. The random forest model showed that the importance weight of canopy closure (0.76) surpassed that of flight altitude (0.24), indicating that canopy structure was the dominant factor. This study clarifies the spatial adaptation boundary of UAV-based thermal infrared recognition and provides a theoretical basis for flight strategy formulation and sensor optimization in early forest fire monitoring.