In vehicular edge computing (VEC) environment, vehicles can offload computationally intensive tasks to VEC servers. However, the high-speed mobility of the vehicles and the uneven distribution of vehicles accessing the servers lead to a low success rate of the task upload and an unbalanced workload among the different VEC servers, which in turn affects the quality of service (QoS). A probabilistic algorithm based on alternating direction multiplier method (ADMM) was designed to solve the task upload problem of vehicles, which improved the success probability of task upload. In addition, an edge-cloud collaborative load balancing task offloading scheme was designed with the aim of minimizing the offloading delay to balance the load among servers through edge-cloud collaboration, thereby improving the QoS. Simulation results show that this paper has lower offloading latency compared to the benchmark scheme; and the scheme improves the task upload success rate compared to other similar schemes with the same vehicle arrival rate. It also reduces offloading latency by about 20%.