“Natural disasters occur frequently in Yunnan Province, China, causing huge losses of life and property to the local people. In order to carry out disaster relief and rescue more effectively, experts have proposed a target detection technology based on high-resolution remote sensing images of unmanned aerial vehicles and deep learning to quickly locate damaged buildings. In the field of damaged building detection, there are currently two major challenges: firstly, high-resolution earthquake damaged building data is scarce and expensive; Secondly, the small differences between the target to be detected and the background and other features can easily lead to false positives. To overcome these issues, the expert constructed a large-scale high-resolution dataset of earthquake damaged buildings based on drone remote sensing images, covering 4598 remote sensing images and annotating the target buildings in multiple forms. At the same time, the expert also proposed a real-time detection model for earthquake damaged buildings, which incorporates a target feature alignment module, a feature difference calculation module, and a position box detection module with target boundary constraints. After verification, the model achieved an accuracy of 86% on the seismic building detection dataset and achieved good application results in actual scenarios at different locations. This research achievement not only provides new technological means for disaster relief and rescue, but also opens up new directions for the application of drone remote sensing images in disaster monitoring.”