Unmanned aerial vehicle image geological disaster monitoring method based on artificial intelligence

Abstract

This invention discloses an artificial intelligence-based UAV imagery method for geological disaster monitoring, specifically relating to the fields of artificial intelligence and computer vision. It addresses the challenges of robustly identifying and locating small-scale, weak-contrast precursors such as newly formed cracks, shallow bulges, and fine debris accumulations under conditions of heavy rainfall / earthquake-affected mountainous areas, limited airborne computing power and data links, and scarce control points, even under interference from rolling shutter speeds, attitude fluctuations, and shadow occlusion. The method employs micro-baselines created by applying micro-jitter under unified timing coordinates, event-driven narrow-band precision detection, and attitude covariance-gated fusion; row-level rolling shutter compensation and pose tight coupling; and skeleton neighborhood parallax to extrapolate scale, distance, and contact time. Unified uncertainty is used to generate confidence ellipses and evidence chains, which are then reported hierarchically. This achieves high recall and low false alarm rates for small precursors, minute-level image generation, and sub-meter-level localization even with few or no control points and limited data link computing power.

Images