The rapid development of unmanned aerial vehicle (UAV) technology has greatly expanded the application scope of the Internet of Things (IoT). Particularly, within forest fire scenarios characterized by the absence of communication infrastructure, the lifespan of ground nodes becomes unpredictable, and the operational environment for UAV becomes increasingly challenging and complex. It is imperative to adopt a trajectory planning methodology for UAV data collection missions, which not only facilitates prompt data acquisition but also ensures the operational safety of the UAV. In response to these challenges, this paper introduces the concept of unit data value for nodes and presents a trajectory planning algorithm for UAV data collection, grounded in the Double Deep Q-Network (DDQN) framework. The primary objective of this algorithm is to maximize the value of the data gathered during the UAV's flight cycle, while simultaneously ensuring the promptness of data acquisition and optimizing the UAV's flight trajectory for safety. In comparison with the baseline algorithms, namely Dueling DQN and DDQN, the novel approach presented in this study demonstrates substantial enhancements in performance across multiple dimensions, including convergence speed, overall data collection efficiency, success rate in planning safe trajectories, and rate of collecting high-value data.