The paper introduces the research and practice of multimedia services of fine-grained object tracking systems infrastructure based on cloud-edge collaboration. Firstly, we pointed out the disadvantages of the basic form of traditional centralized deployment of object tracking systems in the cloud, analyzed the challenges posed by high-performance systems. Secondly, the deployment of the cloud-edge collaborative object tracking system proposed in this paper is introduced, the task segmentation and coordination form between cloud-side and edge-side will be discussed in detail. Finally, we verified the effectiveness of the optimized deployment scheme in this paper by comparing the load-bearing capacity and time delay using the same deep learning model with the centralized deployment scheme.