To address the problem of heavy computational tasks and uneven resource utilization among devices in multi-point monitoring video analysis in intelligent transportation, a cloud-controlled video analysis offloading scheme is proposed. Firstly, for the problem of insufficient client computing power to complete video analysis tasks, a video offloading framework is used to segment and offload some of the video analysis tasks to cloud servers. Then, a stage-optimized offloading algorithm is proposed to balance the load between servers and multiple clients, and improve resource utilization. Finally, to address the issue of different client requirements at different points, precision and energy consumption preferences are added to the algorithm to meet the needs of different clients. Experimental comparisons with other offloading schemes demonstrate that this scheme can better allocate video analysis tasks, improve system benefits, and the scalability of the system is demonstrated through extended experiments.