The large-scale application and intensive flight of Unmanned Aerial Vehicle （UAV） in the future bring about safety problems. In view of the complex low-altitude system formed by the interweaving of diversified airspace， high-density traffic and high dynamic evolution， the problem of network modeling and refined management of complex low-altitude UAV flight conflicts is studied. Firstly， the Gilbert-Johnson-Keerthi （GJK） algorithm is used to construct a UAV flight conflict network model based on the dynamic identification of conflicting edges， and the characteristic parameters of the UAV flight conflict network are characterized by the average number of conflicts， the average operation risk and the average clustering coefficient. Then， based on the spatial raster coding theory， a raster segmentation and coding design method for low-altitude airspace system is proposed， and a digital raster-based UAV flight conflict detection algorithm for discrete airspace environment is established. Finally， a multi-objective mixed integer optimization model for UAV conflict resolution considering priority is constructed with the optimization objectives of the minimum number of UAV conflicts and the lowest operation cost. Moreover， an intelligent UAV conflict resolution algorithm based on Non-Dominated Sorting Genetic Algorithm-Ⅱ（NSGA-Ⅱ） algorithm is designed to realize the optimal resolution of UAV flight conflicts under multi-task scenarios and different priorities. The optimal resolution of UAV flight conflicts under multi-mission scenarios and different priorities is then achieved. Simulation experiments show that compared with the traditional 3D coordinate-based conflict detection method， the proposed digital raster-based conflict detection method can reduce the total detection time by 78.4%， and reduce the UAV conflict detection efficiency from “exponential growth level” to “linear growth level” in multi-model hybrid scenarios. When the average conflict risk level is used as the deconfliction priority principle， the UAV conflict deconfliction model can obtain non-dominated solutions with better performance and reduce the average adjustment values of altitude layer， horizontal trajectory and speed per UAV by 32.8%， 21.4% and 14.6% compared with the principle without considering the priority. It is verified that the fine management method of UAV flight conflict proposed in this paper is effective and can provide theoretical support and methodological guidance for the safety management of low-altitude airspace and UAV risk prevention and control.