Unmanned aerial vehicles (UAVs) are widely used in reconnaissance and environmental monitoring. To improve heterogeneous multi-UAV multi-region coverage in complex terrain, this paper proposes a hierarchical cooperative optimization framework. A task-allocation model minimizing mission makespan is first formulated with sensing-resolution matching, endurance, and terrain clearance as hard constraints. Adaptive Collaborative Artemisinin Optimization (ACAO) then searches the UAV assignment and region-visiting sequence using constraint-based initialization, adaptive updates for the region-visiting sequence, and Linnik-distribution-based differential perturbation. For each decoded route, an endurance-aware routewise entry–exit and replenishment repair strategy, termed Look-Ahead-2 (LA-2), evaluates the current region and its immediate successor, selects among 12 legal ordered vertex pairs, and inserts return and replenishment operations when required. Entry–exit-constrained Boustrophedon coverage and terrain-aware transition repair provide route costs. Experiments on four terrain scenarios with 30 independent runs show that ACAO reduces the mean mission makespan by 13.84%–26.45% relative to the best baseline in each scenario. LA-2 also achieves solution quality comparable to deeper look-ahead strategies and fixed-route exact dynamic programming at a lower runtime.