Abstract: The ever-expanding applications of unmanned aerial vehicles （UAVs） have made ensuring low-altitude airspace security and providing reliable communications support for cooperative UAVs a critical challenge. Traditional UAV surveillance technologies， such as radar， radio frequency， and acoustic monitoring， exhibit inherent limitations in deployment cost and coverage range. To address these challenges， integrated sensing and communication （ISAC） technology offers a new avenue for UAV surveillance， leveraging the advantage of reusing the existing communications infrastructure. This paper focused on the precise three-dimensional （3D） localization， velocity estimation， and dynamic communication link optimization of low-altitude UAVs. Specifically， a multibase station （BS） collaborative method was proposed for joint localization and communication beam prediction. First， at the single-BS signal processing level， we performed parallel factor decomposition on the high-dimensional received signal tensor to obtain preliminary estimates of UAV parameters， including range， angle， and Doppler shift. Then， to jointly utilize observational data from different BSs， a multitarget association strategy based on the spatial proximity criterion was designed. This strategy combined the consistency of the UAV target’s 3D position with the inherent automatic parameter pairing property of single-BS tensor decomposition， which enabled the association of velocity information from different BSs and the calculation of the 3D velocity vector of the UAV target. Finally， to optimize the communication link， a dynamic beamforming mechanism based on motion state prediction was proposed. This mechanism enabled precise beam steering by continuously predicting the 3D position of the UAV， thereby enhancing the communication link quality between the BS and high-speed UAV. Simulation results demonstrated that the proposed method achieved high UAV localization accuracy and significantly enhanced communication performance through multi-BS collaborative dynamic beamforming.