A method and system for intelligent evaluation of road diseases based on unmanned aerial vehicle images

Abstract

The application provides a road disease intelligent evaluation method and system based on unmanned aerial vehicle images, and belongs to the technical field of artificial intelligence. Specifically, the method comprises the following steps: collecting road disease images by using an unmanned aerial vehicle, screening and labeling the images to obtain original unmanned aerial vehicle images; performing data preprocessing and adaptive enhancement on the original unmanned aerial vehicle images to obtain corrected unmanned aerial vehicle images; performing multi-scale feature extraction and structure perception feature mining on the corrected unmanned aerial vehicle images to obtain fused features; introducing a type discrimination and risk perception classification mechanism to construct a risk perception disease classification model; training the model to obtain a trained risk perception disease classification model; and using the trained risk perception disease classification model to perform inspection on an actual road to obtain road disease classification results, related classification confidence results and maintenance attention degrees. The application can significantly improve the boundaries and details of disease areas, and enable the model to have higher category judgment and maintenance priority reference value.

Images