In this study, we designed a ground-aerial robotic system for the warehouse inventory. The system fully exploits the advantages of UAV and UGV, giving it a stronger competitive edge than previous solutions. The sensors and hardwares used in this system also have modest requirements for computational resources and electrical power, which allows for long-term inventory. An integrated localization framework is established to achieve an accurate localization of this system. We proposed a LiDAR-Inertial Odometry and a target-based relative localization method to provide accurate pose estimation for the entire system. We conducted a series of tests using motion-capture cameras to validate the accuracy and robustness of the ground-aerial robotic system. In addition, we constructed inventory tasks in a real warehouse to demonstrate the effectiveness of our system. The experimental results confirm that the proposed system achieves comparable working efficiency and localization accuracy in the warehouse inventory.