During the process of autonomous vehicle driving, it is not only necessary to complete motion planning and achieve pedestrian recognition, but also to implement higher precision multi-target tracking. A pedestrian recognition and multi-target tracking algorithm integrating multiple sensors is proposed to address the issues of slow response speed and poor target tracking accuracy in autonomous vehicle driving. Using the Lattice algorithm for path planning, the optimal driving trajectory is obtained through loss function and collision detection. The obstacle position detected by the sensor is converted to the global coordinate system, and Gaussian distribution is drawn on the global grid map. The visible pedestrian is initially determined through threshold. And a multi-objective tracking occlusion processing algorithm is designed based on the detection and tracking strategy to achieve motion estimation of occluded targets in autonomous vehicles. Quantitative, qualitative, and ablation studies on the multi target tracking challenge dataset validate the effectiveness of the algorithm. The experimental results show that the algorithm can accurately estimate the target motion during occlusion and generate complete, high-quality motion trajectories.