To address the diverse practical flight requirements of low-altitude UAVs (unmanned aerial vehicles) in urban environments, a three-dimensional trajectory planning method based on multi-attribute grid features is proposed. First, safety risk, noise sensitivity, and privacy sensitivity were designed as primary grid features. A multi-attribute grid feature evaluation model and a risk assessment model were established, with calibrated feature weights and risk evaluation methods. Subsequently, the safety risk layer was constructed by integrating population density, shelter coverage, and obstacle layers. This layer is further fused with noise-sensitive and privacy-sensitive layers to generate a fine-grained, integrated multi-attribute grid map with a resolution of 30 m×30 m×70 m. Finally, the bidirectional A* algorithm was employed in conjunction with the multi-attribute grid map for initial trajectory planning and smoothing. Experimental results demonstrate that, com-pared to the traditional unidirectional A* algorithm, the bidirectional A* algorithm significantly reduces computation time and total trajectory distance, proving more suitable for UAV trajectory planning. The application of third-order Bezier curve smoothing reduces turning points by over 85%. Compared to methods without multi-attribute grid features, the proposed approach increases high-performance trajectory points by 56.3% with only a 4% increase in path length, achieving substantial improvements in average trajectory performance. Additionally, comparative experiments on diverse mission objectives provide differentiated trajectory recommendations. The research outcomes offer theoretical foundations and technical support for urban low-altitude UAV trajectory planning and the development of related flight management systems.