An air-ground collaborative mobile edge computing (MEC) system composed of flexible and maneuverable unmanned aerial vehicles (UAV) and ground base stations can effectively guarantee the MEC service quality of the Internet of Vehicles for highly mobile users. In order to adapt to the real-time and random variation characteristics of the positions of IoV users, this paper studies the optimization of communication and computing resource allocation in an online air-ground collaborative MEC system. This research divides the flight time of UAV into multiple equal-length time slots. At each current time slot, this paper employs a Kalman filter-based method to predict the positions of vehicle users in the next time slot. Subsequently, based on the resource allocation in the current time slot, we study a problem of jointly optimizing the communication bandwidth, computing task offloading ratio, UAV trajectory, and computing resources of the system to minimize the communication-computation delay of vehicle users in the next future time slot. To efficiently solve this non-convex optimization problem, this paper proposes an efficient alternating optimization algorithm. We decompose the original problem into sub-problems of bandwidth allocation, computing task offloading ratio allocation, UAV trajectory optimization, and computing resource allocation, and solve these sub-problems by alternating iteration. The simulation results verify that the proposed online optimization algorithm exhibits excellent real-time performance. It can significantly reduce the communication and computation delay required for vehicle tasks, demonstrating the importance and feasibility of real-time optimization of communication and computing resources in the connected vehicle network.