Abstract: Dealing with the problem that high-quality expert experience or accurate model-based calculation is not always available, a novel human-machine collaborative damage assessment method is proposed that supports multiple levels of collaboration. Based on the coactive design theory, a damage assessment task is decomposed into several subtasks that can be solved throughanalyzing reconnaissance images taken by an UAV orthrough approximate model-based computing. According to the task requirements as well as the capabilities of human and machine, autonomy levelsare selected accordingly. Finally, the results of image change detection, model simulation and expert assessmentare weighted. The effectiveness of the proposed method has been verified in an experiment, which has been proved to improve the assessment accuracy while reducing the workload of the human operator.