To address the urgent demand for low-altitude collaborative surveillance technology in future intelligent air traffic development, an improved CI (covariance intersection) fusion algorithm based on a cross-covariance compensation mechanism was proposed. By incorporating cross-covariance information into the traditional CI fusion framework, the conservatism in estimating fused error covariance matrices was effectively reduced. Combined with UKF (unscented Kalman filter), a dual-layer filtering fusion framework was established to integrate surveillance data from ADS-B, 5G data links, and BeiDou short messages. Experimental results demonstrate that the root mean square errors of fused trajectories are reduced in the longitude, latitude, and altitude dimensions through the proposed improved CI fusion algorithm, confirming its effectiveness. This approach provides a more accurate and reliable data foundation for target tracking and trajectory prediction tasks, thereby enhancing the safety and operational efficiency of low-altitude airspace.