Integration of UAV Photogrammetry and AI-based Image Recognition for Investigating Occupation of Irrigation Public Lands

Abstract

This study presents an innovative approach combining Unmanned Aerial Vehicles (UAVs) and AI-based image recognition to investigate unauthorized occupation of irrigation public lands. Traditional land survey methods are labor-intensive, time-consuming, and often constrained by inaccessible terrain. By integrating UAV-based photogrammetry and deep learning techniques—particularly the YOLOv7 object detection framework—this research enables efficient acquisition of high-resolution orthophotos and automatic detection of land use violations. Case studies from southern Taiwan demonstrate improved survey accuracy and significant time savings compared to manual inspection. This method provides a scalable, repeatable framework for land resource monitoring and supports more transparent and standardized land management operations.