Disclosed are a method of inspecting and evaluating a coating state of a steel structure, and a system therefor. A plurality of vision images and thermal images are acquired. While acquiring the thermal images, a desired region is heated. After the thermal images and the vision images in a dynamic state are reconstructed into a time-spatial-integrated thermal image and a time-spatial-integrated vision image in a static state, respectively, an overlay image is generated by overlaying the two images. A deterioration region of a coating is detected, and coating deterioration is classified by characteristics. A size of the coating deterioration region is quantified. A thickness of the coating is inspected by analyzing thermal energy measured from the time-spatial-integrated thermal image. A coating grade is calculated by comprehensively evaluating a coating deterioration inspection result and a coating thickness inspection result. A state evaluation report for the steel structure is automatically created.