Unmanned Aerial Vehicle (UAV) path planning has increasingly become the key research point for civilian drones to expand their use and enhance their work efficiency. Focusing on offline derivative algorithms, represented by Rapidly-exploring Random Trees (RRT), are widely utilized due to their high computational efficiency. However, deploying these offline algorithms in complex and changing disaster environments presents its own drawbacks, such as slow convergence speed, poor real-time performance, and uneven generation paths. In this paper, the Artificial Potential Field -Improved Rapidly-exploring Random Trees (APF-IRRT*) path-planning algorithm is proposed, which is applicable to disaster relief UAV cruises. The RRT* algorithm is adapted with adaptive step size and adaptive search range coupled with the APF algorithm for final path-cutting optimization. This algorithm guarantees computational efficiency while giving the target directivity of the extended nodes. Furthermore, this algorithm achieves remarkable progress in solving problems of slow convergence speed and unsmooth path in the UAV path planning and achieves good performance in both offline static and online dynamic environment path planning.