Abstract: At present, the monitoring of illegal land and illegal construction is mostly based on remote sensing images, it can not effectively detect the situation of building addition and construction, and so on. The use of image densely matching point clouds for extracting 3D changes of buildings is an effective way to accurately monitor “two violations”. In Guigang city “two violations” monitoring as an example, using UAV image matching point clouds as the basic data, through the depth of neural network model to extract buildings point clouds, and to detect the change of the different temporal point clouds, after superimposing the management data of natural resources, such as approval and planning, extract the suspected “two violations” spots and monitoring.

Key words: densely matched point clouds, two violations, UAV, monitoring