Application of Artificial Intelligence and UAV Remote Sensing in Regional Vegetation Environment Information Extraction

The presence of land surface vegetation and buildings plays a crucial role in the urban landscape. The point cloud data collected by LiDAR is an important technical means to rapidly construct regional surface 3D model. However, there are some problems such as high cost and low acquisition efficiency. With AI's rapid progress, new solutions are emerging. In this study, we first collected remote sensing images from unmanned aerial vehicle (UAV) to generate dense image matching (DIM) point cloud data. Aiming at the problems such as incomplete noise point removal and poor algorithm universality in the current extraction process of vegetation and buildings, we adopted emerging AI technology to build a multi-feature joint extraction model. The model's effectiveness was demonstrated on an open dataset, showing its ability to separate vegetation from buildings while maintaining 3D extraction efficiency. Finally, we verified the effectiveness of this model with a case study of Guizhou University. This study aims to combine low-cost UAV remote sensing images and artificial intelligence technology to achieve efficient and high-precision extraction of vegetation and building models in a man-machine collaborative way, and provide application potential for realizing regional environmental monitoring related work quickly and effectively.